Precision combination therapies based on recurrent oncogenic co-alterations

Abstract

Cancer cells depend on multiple driver alterations whose oncogenic effects can be suppressed by drug combinations. Here, we provide a comprehensive resource of precision combination therapies tailored to oncogenic co-alterations that are recurrent across patient cohorts. To generate the resource, we developed Recurrent Features Leveraged for Combination Therapy (REFLECT), which integrates machine learning and cancer informatics algorithms. Using multi-omic data, the method maps recurrent co-alteration signatures in patient cohorts to combination therapies. We validated the REFLECT pipeline using data from patient-derived xenografts, in vitro drug screens, and a combination therapy clinical trial. These validations demonstrate that REFLECT-selected combination therapies have significantly improved efficacy, synergy, and survival outcomes. In patient cohorts with immunotherapy response markers, DNA repair aberrations, and HER2 activation, we have identified therapeutically actionable and recurrent co-alteration signatures. REFLECT provides a resource and framework to design combination therapies tailored to tumor cohorts in data-driven clinical trials and pre-clinical studies.

Introduction

Targeted therapies such as HER2 and PARP inhibitors have significantly improved clinical outcomes of cancer patients^1,2,3. Despite the transformative advances in sequencing technologies and precision medicine, durable responses to genetically-matched targeted therapies have remained limited^4,5. The limited success can be at least partially attributed to treatment with mono-therapies targeting individual oncogenic aberrations. Oncogenesis, however, is driven by co-occurring aberrations that lead to co-activated oncogenic processes. A rich repertoire of preclinical and clinical studies has demonstrated that rational combination therapies blocking multiple oncogenic processes are more effective than mono-therapies^6,7,8. Discovery of combination therapies that may benefit large patient cohorts, however, is a daunting task due to the complexity of the actionable oncogenic aberration landscape^9,10.

There is growing evidence that targeting the co-occurring oncogenic driver aberrations may induce effective and durable therapeutic responses. For example, co-alterations in the PI3K-AKT pathway and BRAF impact response to BRAF inhibitors and establish a rationale for combination therapies in melanoma¹¹. Similarly, co-alterations in KRAS, BRAF, and PIK3CA modulate response to EGFR/HER2 targeted therapies in colorectal cancers⁸. Genomic studies focusing on co-driver events in KRAS mutated lung cancers¹² and recent combination therapy trials (e.g., I-PREDICT) focusing on co-alterations also strongly support this argument⁶. Despite the clinical advances, the precision therapies that can target co-alterations remain mostly unexplored. Bioinformatics approaches that extract clinically useful knowledge (e.g., recurrent and actionable co-alterations) from patient data will accelerate the implementation of precision combination therapies.

Bioinformatics and machine learning methods have demonstrated potential for significantly improving precision medicine. A recently developed method, SELECT, matches patient transcriptomes to targeted therapies based on drug-gene synthetic lethality interactions inferred from CRISPR/RNAi screens in cell lines¹³. The method has demonstrated highly promising outcomes for patient selection in clinical trials mainly for mono-therapies. The method DrugCell has generated successful predictions of drug responses in pre-clinical models and stratification of patients that may benefit from CDK4 and mTOR co-targeting¹⁴. It relies on mutational state and drug chemical structure neural networks trained with responses to hundreds of drugs in thousands of cell lines. The OncoTreat method aims to infer protein activity based on mRNA expression data to prioritize drugs that target aberrant protein activities¹⁵. Logical models (e.g., CellNOpt) predict unseen drug responses based on both signaling database information and drug response data¹⁶. We previously developed network modeling methods that nominate precision combination therapies through analysis of the therapy response mechanisms based on proteomics data^17-19. Despite the significant contributions to automating precision therapy selection, a majority of the approaches suffered from one or a combination of shortcomings such that they (i) benefit from a single data type (e.g., mutational or transcriptional) limiting the coverage of alterations, (ii) focus predominantly on mono-therapies, (iii) require comprehensive training datasets, especially for drug responses that are costly to generate in the clinical setting, or (iv) cannot be immediately scaled to a large number of samples due to immense data requirements.

Here, we tested the hypothesis that combination therapies that are tailored to recurrent co-alteration signatures lead to significant improvements in clinical and pre-clinical outcomes. We have developed the machine learning algorithm and precision oncology resource, REFLECT. First, the REFLECT method generates multi-omic co-alteration signatures within a patient cohort (e.g., cohort of EGFR-mutated patients). The REFLECT-generated signatures are a collection of recurrent co-alterations which can be in the form of mutational, copy number, RNA or protein expression alterations. Each REFLECT signature defines a patient sub-cohort whose members are characterized and discriminated from other sub-cohorts by the recurrent co-alterations. Second, REFLECT maps the recurrent signatures to potential combination therapies based on precision oncology knowledge bases^20,21. We validated the REFLECT method in vivo, in vitro, and in clinical settings. We have demonstrated use cases in patients with immune checkpoint therapy (ICT) response markers, DNA repair aberrations, and HER2 activation. We provide an online resource for > 200 patient cohorts as well as a software package for customized analyses. We expect that the tool and the resource will guide the discovery of combination therapies and the selection of patients who may benefit from combination therapies.

Results

REFLECT-based signatures of oncogenic co-alterations

We have generated a map of the recurrent DNA, mRNA, and protein co-alteration signatures across diverse cancer types using the REFLECT pipeline (Figures 1A-C, S1A-B). The analysis benefited from mutation, copy number, mRNA expression, and phosphoproteomic analyses of patient tumor tissues, cell lines, and patient-derived xenografts (PDXs). The data covers 33 human cancer types (TCGA)²², cell lines from 34 lineages (CCLE, COSMIC cell lines project (COSMIC-CLP), MDACC cell line project (MCLP)) ^23-25 and PDXs from seven cancer types (NIBR PDXE)²⁶ (Figure 2A, S2A-B, Table S1). To correct for batch effects across diverse cancer types, the TCGA protein and mRNA expression data had been integrated and batch normalized with replicate-based normalization (RBN) and EB++ methods^27,28. We also incorporated data from PDXs and cell lines using a Z-score-based normalization (see Methods). The batch effects were not observed across cancer lineages and sources (patients, cell lines, PDXs) within the compendium of mRNA and protein expression data as assessed through principal component analysis (PCA), and uniform manifold approximation and projection (UMAP) analysis (Figure S2C-D). To account for the differences in exome design, depth coverage, and other batch effects in the sequencing data, we used the multi-site, multi-disease, consensus pan-cancer MAF file generated by the TCGA Multi-Center Mutation Calling in Multiple Cancers (MC3) project²⁹.

Figure 1. A. The REFLECT pipeline.

REFLECT identifies significantly recurrent, and therapeutically actionable co-alterations within patient cohorts. B. The integrated pan-cancer data is partitioned into overlapping cohorts based on the existence of potentially actionable genomic aberrations (e.g., cohort of EGFR-mutated tumors). The mutational, copy number, mRNA, and protein datasets from each partition are analyzed separately. C. The significantly recurrent and actionable co-alterations are selected. To identify sub-cohorts of cases characterized by a small set of discriminant features (i.e., genomic alterations), the algorithm iteratively couples LASSO-based feature selection with hierarchical clustering. The significantly recurrent features within each sub-cohort are selected through a statistical assessment (Supplementary Methods). The co-actionable targets are selected by matching features to drugs using precision oncology knowledge bases (e.g., OncoKB and MDACC knowledge bases).

Figure 2. The REFLECT Resource.

A. A pan-cancer meta-cohort is generated from the integration of cancer patient samples (TCGA), PDXs (NIBR PDXE), and cell lines (CCLE, COSMIC-CLP, and MCLP)^22-26 to enable proof-of-principle pre-clinical validations for matching drug combinations to co-alteration signatures. B. The pan-cancer meta-cohort is stratified into 201 cohorts based on the existence of actionable master biomarkers, which are associated with diverse oncogenic processes as demonstrated with a functional GO-term enrichment analysis. C. The summary of the REFLECT-generated actionable and recurrent co-alteration landscape. Each point represents a target for a potential combination therapy matched to a set of cancer patients based on the co-alteration signatures. The heatmap convolutes the results from the DNA, mRNA, and protein analysis. The color code represents the degree of recurrence across cancer patients (most recurrent: red) and hence the size of potentially benefiting patient cohorts. The detailed maps of co-alterations and matched patient cohorts, including genomic variant and cancer types, are provided for each stratification master biomarker in the REFLECT portal.

We partitioned the pan-cancer meta-cohort into patient cohorts, each of which is defined by the presence of a “master” biomarker that is shared by all members of the cohort (Figure 1B, Table S2). The master biomarkers, which stratify the meta-cohort into cohorts prior to REFLECT analyses, are potential therapeutically actionable genomic alterations (e.g., HER2 amplification). The therapeutic actionability of an alteration is defined in reference to the MDACC Precision Oncology Knowledge Base (N=182) (Table S3) and NCI-MATCH precision therapy trials (N=15) (Table S4) ^20,21,30. We have defined and added an immunotherapy cohort based on PD-L1 overexpression and immune cell infiltration (Table S5). Additional cohorts were constructed based on the presence of key oncogenic alterations, MYC amplifications, and TP53 and KRAS mutations (Table S5). In total, we generated 201 master biomarker-based cohorts, which form the core of the REFLECT resource, and yet can be expanded with new cohorts (Figures 1B, 2B). According to OncoKB, at least 25 of the master markers are recognized by FDA and another 14 are recommended by NCCN as predictive of response to targeted therapy in the clinic. The remaining markers are under evaluation in precision oncology trials^20,21.

For each of these 201 cohorts, each defined by a master marker, we generated the multi-omic/multi-gene signatures, termed REFLECT-signatures. In order to capture the co-alteration signatures relevant to cancer and therapeutic targets, we focused on the genes listed within the COSMIC Cancer Gene Census (CGC)³¹ for the genomic and transcriptomic alterations. For signature detection, we use a sparse hierarchical clustering algorithm³² followed by statistical assessment of recurrent alterations (Figure 1C). The sparse hierarchical clustering algorithm iterates LASSO-based feature detection and hierarchical clustering until convergence to a final signature (see Methods). In the succession of iterations, the feature detection selects a small set of alterations that contribute most to discriminating patient sub-cohorts, which are identified by the clustering. Next, the significantly recurrent events from the signature are extracted with a statistical test (see Supplementary Methods). The statistical test filters out the features whose observed recurrence is likely by chance in a large patient cohort. The test preserves the ones that are significantly recurrent and thus potentially functional. The resulting signature carries a combination of the master biomarker (i.e., the identifier of the associated cohort), and a small set of REFLECT-detected recurrent features with non-zero contribution (W) to patient classification. The distinct signatures discriminate sub-cohorts within a patient cohort. To demonstrate the robustness of REFLECT signatures, we quantified the concordance between the features that were calculated with complete versus subsampled data with varying coverage (60% to 95%) (see Supplementary Methods, Figure S2E). The concordance between partial and complete datasets had an average of 83% for 70% coverage across the 7 cohorts tested. The concordance increased monotonically with increased coverage. In summary, we have acquired a compendium of robust multi-omic REFLECT signatures spanning patient cohorts and sub-cohort-specific co-alterations selected by our algorithm.

Landscape of recurrent and therapeutically actionable co-alterations in cancer

Based on the REFLECT signatures, we built a comprehensive resource of recurrent co-alterations that can guide selection of combination therapies tailored to complex molecular signatures (Figure S2F and S2G). For this purpose, we extracted recurrent and co-actionable alterations from REFLECT signatures of 201 patient cohorts. Next, we identified the patient sub-cohorts that carry potentially co-actionable events based on the recurrent co-alterations. The co-actionable targets include (i) the master biomarker that defines the cohort and/or (ii) REFLECT-selected recurrent and therapeutically actionable co-alterations.

To select actionable mutation and copy number events in sub-cohorts, we matched the oncogenic co-alterations in the signatures to the MDACC and OncoKB precision oncology knowledge bases. To match protein and mRNA expression aberrations to therapeutic agents, we developed a preclinical resource based on the drug target information from the MDACC Precision Oncology Knowledge Base, OncoKB, GDSC, and CTRP drug-target annotations (see Supplementary Methods, Tables S6 and S7)^20,21,23,24. The mRNA/proteomics-based actionable events are selected to match drugs to expression aberration of their prospective targets. For example, the increased phosphorylation and total levels of signaling molecules in key pathways (e.g., AKT/PI3K, RTK, MAPK, cell cycle, DNA repair) are linked to their well-characterized inhibitors. Our resulting proteomics- and transcriptomics-based resource matches 82 proteomic and 89 mRNA aberrations to targeted therapy options (Tables S6-S7).

We quantified the number of REFLECT-matched precision combination therapies and recurrent co-alteration signatures (P-recurrence<0.05). Each signature-matched combination involves at least one FDA-approved agent for treatment of one or more cancer (sub)types. The selected drug combinations target a diverse battery of potentially actionable multi-omic alterations such that no individual event dominates the multi-omic target space (Figure S2H). As expected, the most frequently altered oncogenes (e.g., PIK3CA), were among the most common targets (e.g., targeting the PIK3CA alterations is included in 6% of all combinations). The mutation and copy number co-alteration signature analysis led to 325 unique drug combinations. Based on mutational and copy number signatures, 18% of the patients in the meta-cohort are matched to combination therapies. For protein and mRNA aberrations (∣Z-expression∣ > 2, P-recurrence < 0.05), 1235 and 1087 unique signatures were matched to FDA-approved drugs, respectively. Based on protein and mRNA analyses, 32% and 29% of patients are matched to at least one combination therapy, respectively. A total of 2,166 combinations with one or more FDA-approved agents were matched to mutation/copy number, mRNA, and protein co-alteration signatures. In total, 45% of the patients in the meta-cohort were matched to at least one combination therapy based on the multi-omic data.

The majority of the REFLECT-selected co-alterations involved different genes such that therapeutic strategies to target the co-alterations would likely involve drug combinations. We have detected only 50 co-alterations that converged on the same gene (i.e., a gene in a REFLECT signature is identical to the corresponding stratification marker). Given that there are thousands of actionable co-alterations in the REFLECT resource, such events can be considered rare. Meanwhile, within the same gene co-alterations, we observed an enrichment of events involving concurrent amplification and overexpression of the same gene (60% of the same gene co-alteration cases). In the remaining cases, the mutational alterations usually lead to increased phosphorylation and likely hyperactivation of the corresponding protein products (e.g., mutations in EGFR, BRAF, AKT that led to increased protein phosphorylation). The exceptions were increased KIT or EGFR total levels in tumors that carry KIT or EGFR mutations. We also observed that, in the copy number analysis, nearly 60% of the gene amplification and deletion events were mapped to the same chromosomal regions and resided in the identical GISTIC regions. These observations suggest arm-level CNAs play a dominant role in defining the co-alterations space, however co-alteration of focal chromosomal events residing on distinct locations are still common (40%).

Next, we addressed whether the REFLECT-generated pan-cancer signatures are solely driven by cancer types/cell of origin or tumor proliferation levels as opposed to the landscape of multi-omic alterations (Figure S2I). To assess tumor proliferation levels, we calculated the proliferation scores based on the expression levels of key proliferation genes reported in Ref.³³ (see Methods). We observed a weak correlation (Spearman’s correlation= 0.14) between clustering patterns (i.e., ranking of samples across dendrograms) and proliferation scores. The median concordance between cancer types across the REFLECT-dendrograms was also moderate across all REFLECT analyses (N=201 stratifications) with a concordance score of 0.39 (range: 0 to 1, see Methods). More importantly, the distribution of cancer type concordance scores across all REFLECT analyses follows a bimodal distribution such that for a large fraction of stratified cohorts, the median concordance was low (0.25) while a smaller population of cohorts had a higher concordance (median = 0.6). The results suggest that in a larger fraction of the cohorts, clustering patterns were driven by the distribution of the molecular features that are affected by factors other than tumor type; in a smaller fraction of cohorts, tumor type is critical but likely not the only factor determining the clustering patterns. The analyses suggest the REFLECT signatures are not dominated by cell of origin or proliferation levels of the tumors. In summary, the outcome of the REFLECT analyses is a comprehensive resource of precision combination therapies tailored to recurrent co-alteration signatures (Figure 2C and REFLECT portal for co-alteration maps).

Validation of the REFLECT pipeline in PDX pre-clinical trials

We then tested whether REFLECT-matched drug combinations can induce significant therapeutic benefit in a PDX pre-clinical trial. The pre-clinical trial recapitulates a clinical trial design and captures the interpatient heterogeneity through the “one animal per model per treatment” approach (1 × 1× 1)²⁶. We analyzed the therapeutic effects including therapy efficacy, survival, and synergy of the drug combinations in PDXs. The validations compare the therapeutic effects of REFLECT-matched combination therapies versus the combinations that are not matched by REFLECT.

The in vivo data is available for a large array of PDXs from NIBR PDXE (277 PDXs from 7 cancer types treated with 22 targeted drug combinations in 1168 independent treatments)²⁶. The PDXs are profiled for mutation, copy number alteration, and mRNA expression levels (Figure S2A-B). The accuracy, robustness, and clinical relevance of the resource had been established through replication of the drug response-genotype associations, mechanisms of resistance from clinical studies, and > 2,000 single drug response data points from historical PDX experiments²⁶. The established quality and clinical relevance of the screen suggests that the data is well suited for validation of the REFLECT pipeline.

In the validations, we used a merged dataset of patient and PDX samples to execute the REFLECT pipeline and identify the predictions in PDXs. The mRNA data is normalized with the Z-score-based normalization (see Methods). We calculated the multi-omic REFLECT signatures and matched drug combinations to recurrent co-alteration signatures (Figure 3A, S3A, Table S8). FDA-approved drug combinations were REFLECT-matched to 36% of the PDX models (100 out of 277 PDXs). Fifteen of the REFLECT-predicted combinations had been experimentally tested in matched-PDXs within the NIBR PDXE resource (Table S9). We compared the therapeutic benefit by REFLECT-matched drug combinations (REFLECT(+)) to unmatched combinations (REFLECT(−)) (Figure 3B-D). The median tumor volume change with respect to the baseline was 35.4% decrease for REFLECT(+) and 5.1% increase for REFLECT(−) cases (two-tailed t-test, p-value=0.017) (Figure 3B). The progression-free survival (PFS), defined as the amount of time that PDXs take to double their volume from the baseline²⁶, of REFLECT(+) PDXs was significantly longer than that of REFLECT(−) PDXs (Figure 3C for Kaplan-Meier curves and survival statistics). Finally, we quantified the drug synergy using the highest single agent (HSA) model, which provides a simple and unbiased framework to compare drug synergy differences between REFLECT(+) and REFLECT(−) cases. The HSA model quantifies the improvement in efficacy of a drug combination over the agent that is a part of the combination and carries the highest impact as a mono-agent. The drug combinations in the REFLECT(+) cohort had a significantly higher synergy (median HSA score: 28.02) compared to the REFLECT(−) cohort (median HSA score: −0.24) (two-tailed t-test p-value=0.005) (Figure 3D). There is a significant enrichment of synergistic interactions in the REFLECT(+) cohort such that 75% of the REFLECT(+) and 27% of the REFLECT(−) combinations are synergistic (HSA score > 20, Fisher’s exact test p-value=0.003) (Figure S3B). On the other hand, we observed antagonistic interactions in 30% of the REFLECT(−) combinations in contrast to only 8% antagonism in the REFLECT(+) group (HSA score < −20).

Figure 3. The systematic validation of REFLECT.

A. The REFLECT-validation scheme in PDXs (NIBR PDXE) (See Figure S3A for details). The mutation, copy number, and mRNA expression data from PDXs and patients (TCGA) are merged using the z-score based integration method (see Methods). The merged datasets are run through the REFLECT pipeline. PDXs that carry the REFLECT-signatures and are treated with the matched drug combinations are REFLECT(+) whereas PDXs that are not matched by REFLECT to any drug combination are REFLECT(−). The comparison of drug responses in REFLECT(+) vs. REFLECT(−) cases quantifies the benefit from REFLECT. B. The tumor shrinkage in response to drug combinations in REFLECT(+) (N=15) vs. REFLECT(−) PDXs (N = 1153). Note that although 36% of the PDX models (100 out of 277) were matched to at least one drug combination by REFLECT, only 15 of the matched combinations had been tested in the NIBR PDX screen. C. The progression-free survival differences in REFLECT(+) vs. REFLECT(−) cases. D. The drug synergies in REFLECT(+) vs. REFLECT(−) cases. The drug synergies in PDXs are calculated with the highest synergy agent (HSA) model (See Methods for formulation), with higher values corresponding to higher synergy. The statistical significance is assessed with a two-tailed t-test. E. The overlap between the drug combination screens and REFLECT predictions in cell lines covers 315 unique drug combinations spanning 132 REFLECT-matched cell lines (1209 combinations are tested in total in cell lines). The targets designate the altered genes that were targeted in the drug screens, and at the same time, detected by the REFLECT-analysis. The overlapping conditions (REFLECT(+)) enable the systematic validation. F. The comparison of mean synergy scores of combinations targeting the REFLECT-predicted co-alterations in matched cell lines vs. all combinations from the screens (control group). Moderately recurrent: readouts for predictions with P-recurrence < 0.25, Recurrent: P-recurrence < 0.05. Error bars represent standard errors of the mean. G. The predictions from each data modality are compared to the pool of combinations in cell lines. The p-values are based on integration (Fisher’s method) of P-values from t-tests comparing all conditions (moderately recurrent, recurrent, highly expressed) to all screens and corrected for Benjamini-Hochberg FDR = 0.05. H. The clinical validation of the REFLECT pipeline using the I-PREDICT clinical trial data. Patients whose tumors carry the REFLECT-detected recurrent co-alterations and are treated with REFLECT-matched drug combinations are REFLECT(+), and other patients treated with unmatched combinations are REFLECT(−). I. The comparison of progression-free (left) and overall (right) survival of REFLECT(+) vs. REFLECT(−) patients in the I-PREDICT trial.

We have also tested an alternative validation strategy, which involved matching of PDXs to drug combinations using the pre-computed REFLECT signatures based on the TCGA samples only (Figure 2C) and without re-running the REFLECT with the merged dataset (Figure S3C). Compared to the approach that involved re-running REFLECT (Figure 3A), this strategy led to a larger volume for REFLECT(+) cases (N=237) as it lacks the recurrence constraint for the alterations in the PDXs. With this alternative approach, the therapeutic improvement with respect to the REFLECT(−) group is still significant (Figure S3C). As the matching is not constrained by the recurrence criterion, the overall therapeutic gain, however, is moderate compared to the integrated REFLECT analysis with the merged data (Figure 3A-D). In conclusion, the validations in PDX models demonstrated that REFLECT introduced significant therapeutic benefits with improved tumor shrinkage, progression-free survival, and drug synergy.

Validation of REFLECT with multi-omics data and drug screens in cell lines

We devised an independent validation scheme based on three comprehensive drug combination screens in cancer cell lines (Astra Zeneca, NCI ALMANAC, and O’Neil et al resources)^34-36, which had been previously merged and used within the DREAM challenges (Figure S3D)³⁴. The merged screen includes 75,612 unique data points for 3,694 combination targets across 142 cell lines. The mutation, copy number, mRNA expression, and phosphoproteomic data are available from COSMIC-CLP, CCLE, and MCLP projects^23-25. The drug synergy (Loewe) scores and target annotations were already incorporated within the framework of DREAM challenges³⁴. The reproducibility and data quality had been tested through comparisons of replicates across the screens and assessment of curve-fitting for drug responses using the Combenefit analysis suite³⁷. The dataset has already been evaluated and analyzed by > 160 research teams for benchmarking drug response prediction algorithms³⁴. The well-tested drug response resource serves as an unbiased platform for testing REFLECT-predictions with high statistical power.

First, we identified the subset of drug combinations that target the REFLECT-nominated co-alterations in matched cell lines (REFLECT(+)) (Figure 3E, S3D, Table S10). We compared the therapeutic effect as quantified by the synergy scores between the REFLECT(+) combinations vs. all other drug combinations (Figure 3F-G, Table S11). The synergy scores for REFLECT(+) combinations (mean = 25.8, median= 30) are significantly higher compared to scores for all other combinations (mean= 10.5, median= 7.3) (two-tailed t-test, p-value < 0.001) (Figure 3F). For each individual data modality (protein, mRNA, mutation/CNA), REFLECT(+) synergy scores are also significantly higher than the scores for the non-matched cases (Figure 3G, Figure S3E-G, Table S11). The therapeutic benefit monotonically increased with recurrence (low P-recurrence) and expression levels (high Z-scores). The results from the systematic validation scheme suggest that REFLECT-predictions may guide selection of more synergistic drug combinations, regardless of the data modality. Meanwhile, recurrence of co-alterations and protein/gene overexpression are strong predictors of synergistic drug responses.

Validation of REFLECT in the clinical setting

We tested REFLECT in the clinical setting using the patient response and molecular alteration data from the I-PREDICT trial (NCT02534675) (Figures 3H, S3H)⁶. The trial included patients with metastatic cancers (stage IV disease) treated with targeted therapy, in most cases in combinations. The cohort included patients with diverse cancer types, but predominantly having gastrointestinal, gynecological, or breast cancers. The drug combinations were matched to patients based on next-generation sequencing and immunotherapy response markers (PD-L1 IHC and tumor mutation burden) of their malignancies. A tumor board consisting of physicians and scientists matched drug combinations to the response markers. We had access to therapeutically actionable alteration annotations, immunotherapy markers, and survival data of each patient⁶. As our analyses focused on alterations in multi-targets and combination therapies, we excluded monotherapies and combination therapies targeting the same targets (i.e., trastuzumab and pertuzumab targeting ERBB2/3 alterations). We identified patients whose tumors carried REFLECT-determined co-alterations and were treated with the REFLECT-matched combination therapies, the REFLECT(+) cohort (N=12), and patients who did not match to a prediction, the REFLECT(−) cohort (N=17) (Figures 3H, S3H, see Table S12 for the patient characteristics). The treatments in REFLECT(+) and REFLECT(−) groups were highly diverse with minimal overlap. In total, there were 27 unique treatments across 29 patients, with only one treatment administered to two patients in the REFLECT(−) group (trametinib and bevacizumab), and there was only one overlapping combination (trametinib and everolimus) between the two groups. While the drug combinations in both groups involved a total of 25 unique agents, only 8 mono-agents (32% of all agents) were matched in both REFLECT(−) and REFLECT(+) combinations. The remaining 17 mono-agents were administered as part of either REFLECT(+) or REFLECT(−) combinations but not both. The REFLECT(+) cohort had significantly longer PFS and overall survival than the REFLECT(−) cohort (Figure 3I for Kaplan-Meier curves and statistics). There was no observable enrichment of a particular cancer type in either group, suggesting that disease type was not confounding the predictions (Figure S3I). Through a multivariate Cox regression analysis, we validated that the REFLECT status was the only statistically significant predictor of survival against a set of potential confounding factors (PD-L1 expression, tumor mutation burden, and BRCA1 mutation) (Figure S3J-K). MSI status most likely had no impact on the survival predictions as only one patient carried an MSI-high annotation. Thus, the data from the I-PREDICT trial provided strong evidence that REFLECT can provide significant therapeutic benefits in cancer patients.

The PD-L1+ and immune infiltrated tumors are enriched with inflammatory signals and actionable DNA repair pathway alterations

Precision combination therapies that improve responses to immunotherapies (e.g., PD1/PD-L1 inhibitors) are highly desired. Here, we determined the multi-omic signatures and co-occurring actionable targets within cohorts of likely responders to immunotherapies. First, we focused on PD-L1 expression as it has been established as a potential marker of PD1/PD-L1 inhibitor response in select cancers. In addition to the PD-L1 expression, another response predictor is the immune cell infiltration into the tumor microenvironment (TME)³⁸. To establish a robust PD-L1 high cohort, we included the tumors whose PD-L1 mRNA and PD-L1 protein levels are in the top quartile across the pan-cancer meta-cohort. Next, we selected the PD-L1 high tumors whose tumor-infiltrating leukocyte (TIL) fraction is in the top quartile across the pan-cancer meta-cohort (Figure 4A-B). The TIL fraction had already been quantified with a mixture model of tumor subpopulations based on DNA methylation markers with the greatest differences between pure leukocytes and non-leukocyte tissues³⁹. The resulting high PD-L1 and high TIL cohort contains 406 cases from 23 disease types (Figure 4C, Table S5).

Figure 4. REFLECT analysis of the anti-PD1/PD-L1 response marker cohort.

A. Patients whose tumors carry the potential immunotherapy response markers (PD-L1 expression and immune infiltration score) are selected. B. The patient group that corresponds to the intersection of upper quartiles of PD-L1 protein expression, CD274 (coding PD-L1) mRNA expression, and immune infiltration are included. C. The intersection of the cases carrying the markers led to a cohort of 406 patients. D. The REFLECT signatures of the PD-L1/TIL cohort with significantly recurrent features for mutations, mRNA expression, and phosphoproteomics. E. The cellular functions that are significantly enriched in the mRNA REFLECT signature as determined by functional enrichment analysis (WebGestalt enrichment analysis server, over-representation analysis, gene-ontology molecular function database, BH-FDR < 0.05). F. The statistically significant enrichment of naive B-cells and plasma cells in the CCR4 high/TNFSFR17 (CT) high sub-cohorts within the PD-L1/TIL cohort (IO: all cases including CCR4/TNFSFR17 high sub-cohort).

Using REFLECT, we identified the significantly recurrent mutational, mRNA, and phosphoproteomic signatures that co-occur with high PD-L1/TIL (Figure 4D, S4A). The mutation analysis of the PD-L1/TIL cohort identified a signature of eight genes (P-recurrence < 0.05) covering the commonly mutated oncogenes in cancer. The most recurrent signature genes (P-recurrence < 0.0001) discriminating patient sub-cohorts are TP53 (50% of the cohort), PIK3CA (15%), and CDKN2A (10%) followed by RB1, BRAF, and KRAS. The mRNA REFLECT signature is characterized by a combination of significantly recurrent events (P-recurrence < 0.05) in immune genes as well as DNA damage response and repair markers. The strongest contributing mRNA feature is the overexpression of S100A7 (P-recurrence < 0.0001). In the context of cancer, the S100A7 functions in the TME with roles in inflammation-driven carcinogenesis and recruitment of immune cells, including tumor-associated macrophages (TAM). S100A7 is highly expressed in a series of cancers including head and neck cancers^40-42. The S100A7 enriched sub-cohorts also carried increased expression of TP63, HRAS, and FGFR3 particularly in head and neck, bladder, and squamous lung cancers.

The transcriptomic signature is also marked by genes related to DNA repair through chromatin remodeling and epigenetic control (SETD2, NSD1, TRRAP, CHD2, ZMYM3, BRD4), G2-M DNA damage checkpoint (ATM, ATR), and nucleotide excision repair (XPC, ERCC2). We observed two patient sub-cohorts with enrichment of epigenetic regulators of DNA damage response and repair. The first sub-cohort, composed of head and neck cancer cases, is characterized by low expression of the DNA damage checkpoint kinase ATM, as well as epigenetic regulators SETD2 and ZMYM3 genes, which function in recruitment of key enzymes (e.g., BRCA1 and RAD51) in both homologous recombination (HR) and mismatch repair⁴³. A sub-group of this patient group is also characterized by the concurrent loss of nucleotide excision repair (NER) enzyme XPC, and HRR protein XRCC2. The second sub-cohort was composed of patients with diverse cancer types (breast, stomach, lung, and colorectal cancers) and characterized by high expression of epigenetic regulators NSD1 (encoding SETD1 protein) and TRRAP which links chromatin remodeling to DNA double-strand break (DSB) repair⁴⁴. A functional enrichment analysis using the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt)⁴⁵ revealed highly significant enrichment of gene sets involved in histone methylation (FDR-p<0.001), DNA repair-associated processes, replicative senescence (FDR-p=0.03), and telomere maintenance (FDR-p=0.01) (Figure 4E). Another sub-cohort populated by lung squamous cancer cases was characterized by co-expression of the immune chemotactic receptor CCR4 and TNFRSF17, a marker of B-cells^46,47. We demonstrated, based on a CIBERSORT analysis, that tumors in the high CCR4/TNFRSF17 co-expression sub-cohort are significantly enriched with naive and plasma B-cell markers (Figure 4F).

In the phosphoproteomics analysis, depletion of RAD50, a key protein for initial processing of DSB prior to HRR and tethering the two broken ends of damaged DNA, has the highest contribution to discriminating the PD-L1/TIL sub-cohorts (Figures 4D, S4A). The RAD50 protein loss was recurrent in two patient sub-cohorts; one is populated by diverse cancer types (N=111, P-recurrence < 0.0001) while the other is enriched by bladder cancers (N=48, P-recurrence < 0.0001). Increased CHK1 protein level (P-recurrence < 0.0001, N=25) and CHK2 phosphorylation at T68 (P-recurrence < 0.0001, N=52), a marker of DNA damage response and G2-M DNA damage checkpoint, also co-occurred within a fraction of the RAD50 depleted sub-cohort, suggesting aberrant HR-based DSB repair in such tumors.

Alterations in DNA repair pathways co-exist with diverse inflammatory and immunotherapy response markers

We increased the scope and depth of the immune analysis to identify potentially actionable and recurrent alterations that are robustly associated with diverse immunotherapy markers. This effort is based on the rationale that actionable and recurrent alterations that consistently overlap with diverse immune markers may inform potentially effective combination therapies. In addition to the PD-L1/TIL cohort, we defined 10 patient cohorts marked with combinatorial enrichment of potential immunotherapy response markers (tumor immune infiltration, tumor mutation burden, T-cell inflammatory signature), major immune cell types (CD8+ T-cells, CD4+ T-cell, B-cells, Treg cells), and expression of targets of immune checkpoint therapies that are in clinical use or trials (CTLA4, ICOS, TIGIT, LAG3) (Figure S4B)^48-55. The proportion of immune cell types in each patient sample had been inferred by a CIBERSORT analysis of transcriptomics data and imported from Ref.³⁹. The cohorts are generated through the inclusion of tumors whose levels of corresponding markers are in top quartiles across the pan-cancer meta-cohort as in the case of the PD-L1/TIL cohort (Figure 4A-C). To increase the depth of the REFLECT analysis beyond the COSMIC CGC, we expanded our gene set by the addition of the genes in the IFN-γ gene signature, the expanded immune gene signature and the major histocompatibility complex (MHC) – the gene sets that are associated with immunotherapy responses (Figure S4C)⁵¹. We generated the REFLECT signatures and analyzed the mRNA (based on the expanded gene list) and protein signatures across the 11 immune cohorts.

The transcriptomic signatures within the PDL1/TIL cohort remained enriched by mediators of DNA damage response (DDR) when we added the immune related genes to the COSMIC CGC set (Figure 4D, S4D). A gene set enrichment returned DDR and histone methylation among the top significantly enriched processes (Figure S4E). The mRNA co-alteration signatures, however, were also enriched by a set of inflammatory features, most striking being IFNG expression (Figure S4D). In the PDL1/TIL cohort, we observed a highly recurrent and strong IFNG expression signature that spans diverse tumor types (melanoma, sarcoma, bladder, breast and cervical cancers). The IFNG overexpression was mutually exclusive with expression of the inflammatory protein S100A7 (mainly in head and neck cancers) (Fisher’s exact test, P-value < 1.0e-6), while the patients whose tumors express the LAG3 immune checkpoint receptor were overlapping with both IFNG and S100A7 enriched sub-cohorts. LAG3 and S100A7 co-expression also overlaps with expression loss of a set of DNA repair genes such as DNA damage checkpoint kinases (ATR, ATM), telomere maintenance gene POT1, epigenetic regulator ZMYM3 and the base excision repair gene XPC, suggesting DNA repair and inflammatory events co-exist in this sub-cohort. Next, we performed an integrated analysis of the discriminant RNA alterations that were shared across the different immune cohorts (Figure S4F). This analysis generated a consensus signature across the cohorts each marked by a different set of immune markers (immune checkpoint expression, cell type, mutation burden). The consensus signature was highly enriched by the recurrent inflammatory alterations involving the overexpression of genes encoding IFNG, CCR7 and CCR4, B-cell lymphocyte lineage markers PAX5 and CD79A-CD79B (B-cell receptor complex), co-stimulatory immune receptor CD28, and MHC complex.

We calculated the consensus proteomic alteration signatures by quantifying the discriminating features that were shared across different immunotherapy response cohorts (Figure S4G). Consistent with the co-alteration landscape of the PD-L1/TIL cohort, the REFLECT analyses showed that the recurrent loss of the HR repair protein RAD50 was the most discriminant proteomic alteration in nearly all immune cohorts (Figure S4G). Other DNA repair proteins; single-strand break (SSB) repair protein XRCC1, G2M checkpoint kinase p-CHK2, and epigenetic regulator of DNA repair SETD2 were among the most recurrent co-alterations with the immune markers. Finally, although not as strong as the DNA repair and inflammatory signals, we observed signatures associated with activation of the oncogenic growth pathways (e.g., RAF1, p-EGFR, mTOR, AKT, cMET) in the proteomic and transcriptomic modalities (Figures S4F and S4G).

In summary, we have consistently observed co-alterations in DNA repair, particularly the HR repair mechanisms (e.g., loss of RAD50), DNA checkpoint enzymes (CHK1/2, ATM), and epigenetic regulators of DNA repair (e.g., SETD2) in combination with enrichment of diverse inflammation markers. The co-alteration signatures involving DNA repair, oncogenic signaling and inflammation signatures warrant further studies to investigate therapeutic benefits first in pre-clinical models and next in cancer patients, whose tumors carry the therapy matched co-alterations.

Recurrent co-alterations within DNA repair pathways nominate PARP inhibitor-based combination therapies

PARP inhibitors have provided significant therapeutic benefit, particularly for treatment of tumors with DNA repair defects. Here, we have nominated combination therapies in which PARP and other DNA repair inhibitors serve as the anchor compound. We applied REFLECT to patient cohorts marked with alterations in 22 DNA repair genes (Table S13). The selected genes are implicated in BRCAness and potentially synthetically lethal with inhibition of PARP (or other DNA repair enzymes), as listed in the MDACC Precision Oncology Knowledge Base²⁰. The DNA repair genes were selected based on the annotations in MDACC and MSKCC precision oncology knowledge bases with refinement of the BRCAness and PARP inhibitor sensitivity information through literature curation⁵⁶. All of the alterations involved genes that function in DNA double-strand break repair with the exception of the mismatch repair gene MSH3. A majority of the genes plays direct roles in homologous recombination repair (HRR). For example, the genes encoding members of the MRN complex (i.e., RAD51 and MRE11) tether DNA double-strands prior to HRR. BRCA partners such as BARD1 and PALB2 are included. The gene encoding the key enzyme of the HRR mechanism RAD51, which catalyzes the strand transfer, as well as various RAD51 partners (e.g., RAD51B/D, RAD52, RAD54, XRCC2), are included. The list also includes the genes encoding the G2-M DNA damage checkpoint enzymes (ATM, ATR, CHK1/2). Despite being a mismatch repair gene, MSH3 was included in the list as BRCA1 is a downstream effector of the MSH3 and it was included as a BRCAness gene that is potentially associated with PARP inhibitor sensitivity⁵⁶. First, we generated the multi-omic signatures for each of the cohorts to identify recurrent co-alterations. Next, we selected potential combination therapies using the forward target selection criterion (Figure S1B, Table S14).

To achieve an integrated analysis of the signatures associated with the DNA repair alterations, we mapped the proteomic co-alterations paired with cohort markers on a bipartite graph (Figure 5A). On the graph, each recurrent feature within a cohort is linked to the respective co-occurring master biomarker of the cohort (i.e., the DNA repair aberration that defines a cohort) (Figure 5A). We observed an enrichment of aberrant expression in DNA repair proteins (RAD51, ATM, CHK1/2), RTKs (p-EGFR, p-HER2, MET, KIT, VEGFR), and downstream signaling proteins (p- AKT, mTOR, p70S6K, p-JNK, p-SRC, pP38/MAPK14, p-PDK1, p-RAF, p-MAPK) that co-occur with the master biomarkers in the DNA repair family. Guided by the graph enrichment of co-alterations in DNA damage markers, RTKs, and downstream signaling, we experimentally co-targeted the PARP and co-occurring aberrations (Figure 5B-D, S5A-C). In the drug perturbation experiments, we used the cell lines molecularly matched to (i.e., co-cluster with) the specific patient sub-cohorts carrying the co-alterations. We prioritized molecularly matched gynecological and breast cancer patient sub-cohorts because PARP targeting has demonstrated strong, albeit transient, therapeutic benefit for some patients with high-grade serous ovarian or basal-like breast cancer. Using the forward target selection (Figure S1B) in the RAD50, PALB2, and ATM altered cohorts, we identified sub-cohorts characterized by statistically significant recurrent CHK2 phosphorylation, AKT phosphorylation, and MET expression, respectively (P-recurrence < 0.25). For each of these sub-cohorts, we identified a cell line with a matched co-occurrence pattern, namely MFE-296 (RAD50 mutation with recurrent CHK2 phosphorylation), HEC50B (PALB2 deletion with recurrent AKT phosphorylation), and DU4475 (ATM deletion with recurrent MET protein overexpression). The drug combinations produce additive to synergistic effects in each of the tests in the matched cell lines, as quantified by the combination index (Figure 5B-D). We have also tested the same drug combinations in unmatched cell lines that do not carry the corresponding co-alterations (i.e., REFLECT criterion for unmatched: P-recurrence > 0.25) (Figure S5A-C). We did not observe therapeutic gains by the combinations in the negative control cases. Overall, the REFLECT calculations determined the landscape of co-alterations involving DNA-repair aberrations. When coupled with pre-clinical and clinical translation, the landscape may inform selection of effective combination therapies tailored to co-alterations.

Figure 5. Targeting co-alterations in tumors with DNA repair aberrations.

A. The landscape of actionable co-alterations associated with DNA-repair genes. The bipartite graph links the alterations in the DNA repair genes (master biomarkers, green) and actionable protein co-alterations (REFLECT-detected, pink). The node size is proportional to the degree of connectivity that represents the number of co-alterations with the master marker. The edge width is proportional to the number of patients that carry the co-altered nodes. The links between select genes whose alterations were implicated in PARP inhibitor synthetic lethality (MDACC knowledge base )²⁰ and actionable proteins matched to FDA-approved drugs are shown. B-D REFLECT analysis of the mutations/deletions of RAD50 (B), PALB2 (C), and ATM (D), which demonstrate potential synthetic lethality with PARP inhibitors. The top row: weights (w) of discriminant features from sparse hierarchical clustering, the second row: the REFLECT-signature with recurrently altered protein expressions (Z scores) (P-recurrence < 0.25), the third row: a specific sub-cohort with recurrent patterns and matched cell lines, and the bottom row: proof-of-principle validations in the matched cell lines. The cell lines are treated with drugs targeting the co-alterations and profiled for cell viability. The drug synergy is quantified using the combination index (CI) score at IC50.

Combination precision therapies for HER2-activated tumors

Therapeutic targeting of HER2 has improved clinical outcomes particularly in HER2+ breast cancers, yet resistance to therapy remains a major clinical problem. Drug combinations that will improve the therapy response rates among HER2+ tumors are highly needed. We used the REFLECT pipeline and a reverse search strategy (Figure S1B) to identify patient cohorts that may benefit from combination therapies involving HER2 inhibitors. In the reverse search strategy, we scanned proteomic signatures within 201 cohorts, each defined by the actionable biomarkers. We identified patient sub-cohorts carrying actionable oncogenic aberrations (master biomarkers) and HER2 activation markers (increased total protein or phosphorylation (p-Y1248) levels). The analysis led to the identification of 69 sub-cohorts that carry a HER2 activation marker (P-recurrence < 0.05) alongside an actionable master biomarker (Figure 6A). As a result, 338 patients (4% of the TCGA meta-cohort) were matched to 51 unique combinations involving HER2-targeting (Table S15). For proof-of-principle validation experiments, we identified cell lines that map to sub-cohorts of increased HER2 activity. The molecularly matched cell lines were sampled within the cohorts of (i) AURKA mutation/amplification (cell line: BT474), (ii) MAP3K4 mutation/amplification (cell line: SKOV3), (iii) CDKN1B mutation/loss (cell line: HCC1954), and (iv) PTEN loss/mutation (cell line: MDAMB468) (Figure 6B-E, S6A-D). These cell lines were perturbed with HER2 inhibitors and agents targeting the cohort biomarkers. The drug combinations targeting the co-alterations substantially improved responses compared to single agents. The drug combinations carried moderate to strong synergies as quantified by the combination index (Figure 6B-E). The same drug combinations involving HER2 inhibitors did not lead to significant therapeutic gains in the unmatched cell lines that serve as negative controls (REFLECT criterion for unmatched: P-recurrence > 0.25) (Figure S6A-D). More importantly, we have mapped the co-alteration signatures that may guide selection of patients with HER2+ tumors and match to 51 unique drug combinations involving HER2 inhibitors. The REFLECT calculations generated a precision combination therapy resource for the treatment of HER2-activated tumors through targeting co-alterations.

Figure 6. Co-targeting strategies in HER2-activated tumors.

A. The co-alteration landscape of HER2 protein overexpressing cancers. A bipartite graph showing the co-alterations between stratification genes (master biomarkers, green) and HER2/HER2_pY1248 (pink). In the graph, the size of a node is proportional to the degree of the node, and the width of an edge is proportional to the number of patients matching the combination pair. B-E. Co-targeting of HER2 activation with AURKA mutation/amplification (B), CDKN1B loss (C), MAP3K4 mutation/amplification (D), and PTEN loss (E). In each panel, the top row shows weights (w) of features from sparse hierarchical clustering. The second-row displays REFLECT signatures with recurrently up/down-regulated protein expressions (Z scores) (P-recurrence < 0.25). The third row magnifies the region with the selected recurrent pattern (e.g., HER2/HER2_pY1248 upregulation) and matched cell lines in the second row. The bottom row shows proof-of-principle tests in matched cell lines. The cell lines are treated with the combinations of HER2/EGFR inhibitor (neratinib) and inhibitors of AURKA (B), CDK4 (C), P38/MAPK14 (D), and PI3K (E). In the bar charts and heatmaps, only the actionable features with non-zero discriminant weights (w) are shown, except for (D) (MAP3K4) where all the non-zero features are displayed. The drug synergy is quantified using the combination index (CI) score.

Discussion

Here, we present a resource and a bioinformatic analysis method, REFLECT, to match combination therapies to combinatorial aberration signatures in patients’ tumors. The REFLECT method includes a machine learning algorithm for the detection of recurrent co-alterations and an informatics component to match the co-alterations to therapies. With a REFLECT-based analysis, we have generated a resource that contains combination therapies matched to > 10,000 patients across > 200 patient cohorts. Our comprehensive validations using cell lines, patient-derived xenografts, and clinical data demonstrate that REFLECT consistently provides significant therapeutic benefits in both pre-clinical and clinical settings. The REFLECT-predictions led to more desired drug responses in terms of efficacy, synergy, and survival outcome compared to control groups for all data sets (pre-clinical and clinical) and modalities (DNA, mRNA, protein). This is particularly striking given that the predictions rely on signatures in therapy naïve patients or pre-clinical models and are not constrained by any drug response data. The ability to nominate drug combinations based on data from pre-treatment tissues is critical as it suggests REFLECT can be easily implemented to currently used or emerging precision oncology protocols. Based on the consistently superior outcomes from REFLECT(+) cases, we conclude that REFLECT has the potential to facilitate precision medicine applications.

REFLECT is built on a set of core principles in cancer genomics and therapeutics. First, oncogenesis and cancer progression are driven by multiple co-existing driver aberrations^12,57,58. Such co-existing aberrations may be causally coupled to each other, and targeting such coupled events may lead to significantly synergistic, strong responses. Alternatively, co-alterations may drive parallel and relatively independent events without causal links in the same or separate tumor sub-clones. Such independent events may contribute to cancer progression in an additive manner. Subsequently, targeting the independent events may lead to an additive response with increased durability, a highly desired therapeutic outcome. Second, we emphasized recurrence in the analysis of co-alteration signatures. Oncogenic alterations that are significantly recurrent across patient cohorts are likely functional, as such recurrence across samples suggests selection to drive oncogenic hallmarks⁵⁹. The significant recurrence of actionable features also suggests the resulting therapy regimes may benefit substantially large patient populations. Third, most precision therapies are usually matched to functional variants of an oncogene (e.g., BRAFV600E). Here, when selecting potentially actionable mutations, we have not filtered out variants of unknown significance (VUS), instead we have included all variants of a master biomarker if the variant exists in more than two cases in the meta-cohort. This approach involves the risk that some recurring passenger mutations are included in the analysis. This approach, however, has the advantage that we detect the REFLECT signatures that are shared by both known functional/actionable variants and VUS, which may guide therapeutic options to target tumors with important VUS. Finally, the stratification into patient cohorts based on actionable biomarkers emulates the ongoing basket trials such as NCI-MATCH and NCI-ComboMATCH. The approach also carries the added value that it enables matching combination therapies to sub-cohorts with shared molecular compositions. In summary, the success of our approach relies on the observation that recurrent co-alterations that co-activate multiple oncogenic processes may prevent benefit from monotherapies, yet, such co-alterations that are shared across patient sub-cohorts create vulnerabilities to matched combination therapies. The REFLECT resource is highly dynamic and is open to exponential growth with the addition of new cohorts. We anticipate further growth and increased usability of the REFLECT resource as new drug targets are discovered and translated through clinical trials.

To demonstrate the capability of REFLECT, we focused on the signatures of three clinically relevant events: potential immunotherapy response markers, DNA repair/damage response aberrations, and HER2 activation. The analyses of HER2 and DNA repair aberrations have identified patient sub-cohorts matched to the combination therapy candidates that we validate in matched cell lines. The experimental validations reveal either novel drug synergies, such as co-targeting Aurora kinase and HER2 in characterized patient groups, or patients who potentially would benefit from previously tested drug combinations, such as those co-targeting PARP with CHK2⁶⁰, AKT^61,62, or MET⁶³ and co-targeting HER2 with CDK4⁶⁴ or PI3K⁶⁵. Indeed, a large number of synergistic combination therapies have been previously discovered in diverse preclinical studies. It remains elusive, however, if there is a substantial number of cancer patients with a well-defined signature who may benefit from a previously discovered combination. Therefore, a key contribution of REFLECT to future precision medicine applications is in the mapping of recurrent co-alteration signatures to known drug combinations, as well as discovery of novel combination therapies tailored to recurrent co-alteration signatures of cancer patients.

Selection of immune checkpoint therapy (ICT) response markers particularly in a pan-cancer fashion is not a trivial task. Despite their success in select cancers, neither PD-L1 expression nor tumor mutation burden is robust predictive markers across all cancer contexts^66,67. Acknowledging these shortcomings of individual biomarkers, we used a compendium of events involving immune checkpoint expression, tumor immune cell infiltration, and immune cell type proportions to define a set of cohorts that are most likely to be primed for ICT response. The resulting REFLECT-generated signatures of potential response markers involve enrichment of specific inflammation (e.g., IFNG, S100A7, LAG3) and immune cell (B-cells) markers as well as DNA repair alterations, particularly the HR repair and G2-M DNA damage checkpoint mechanisms. The co-enrichment of DNA damage and immunotherapy response markers establishes a rationale for testing PD1/PD-L1 inhibitors (and likely other ICTs) in combination with DNA repair inhibitors in molecularly matched patients. Clinical trials combining PARP inhibitors and immunotherapies are already in progress with promising results in breast, prostate, and ovarian cancers^10,68. However, the therapeutic opportunities are not limited to PARP inhibitors as combinations of ATR and PD-L1 inhibitors have recently progressed to clinical trials⁶⁹. The sub-cohort-specific co-occurrence patterns (e.g., recurrent RAD50 loss) may guide the selection of patients who are likely to benefit and inform which specific DDR-targeting agent may optimally combine with anti-PD1/PD-L1 for a specific indication. Associated with the immune markers, we have also identified a series of co-alterations in epigenetic factors (SETD2, NSD1, TRRAP, CHD2, ZMYM3) that link chromatin remodeling to DNA repair through loading of key enzymes (e.g., BRCA1, RAD51) to damaged DNA^43,44. Preclinical studies are warranted to determine whether loss of such epigenetic factors are synthetically lethal with PARP (or other DDR) inhibitors or alterations involving their overexpression are indicative of therapeutic benefit from bona fide epigenetic inhibitors. An equally important question in the context of immunotherapy trials is on the interactions of such epigenetic alterations with immunotherapy responses. Indeed, a recent study has already suggested that SETD2 mutations are associated with high tumor mutation burden and sensitivity to immune checkpoint therapies possibly through impaired DNA mismatch repair⁷⁰. The immunotherapy opportunities beyond use of anti-PD1/PD-L1 are also compelling. For example, the overlapping LAG3 expression and loss of diverse DDR mechanisms (DNA damage checkpoint, epigenetic regulation and NER) may warrant further investigation of the vulnerabilities to LAG3 targeting in the matched patients. In summary, our studies nominated diverse DNA repair pathway alterations that co-occur with immune activities. Such alterations may guide selection of patients for emerging clinical trials or enable design of novel translational studies with combinations of DNA repair inhibitors and immunotherapies.

The contributions from bioinformatics tools to precision oncology efforts have become more needed as multi-omic data-driven clinical studies (e.g., WINTHER trial) are emerging⁷¹. The clinical implementation of REFLECT for generation of therapy recommendations tailored to patients’ multi-omic signatures is a critical future goal. We expect implementation of REFLECT with an impact on clinical decision making will be possible with the increasing availability of DNA and RNA sequencing data in the clinical setting. DNA sequencing in various forms (e.g., whole-exome and targeted gene panel sequencing) is already in clinical use. Data sets from DNA sequencing across large patient cohorts are immediately amenable to REFLECT analysis. Importantly, REFLECT does not necessitate whole-genome data and could even be implemented with large-scale panel sequencing data. Whole transcriptome data from RNA sequencing is also becoming available in accredited clinical laboratories and its use as predictive biomarkers of response is in progress (e.g., WINTHER trial)⁷¹. Although currently not approved for clinical applications, we anticipate that proteomics will also become a critical component of precision oncology pipelines in the future. This is particularly critical as most targeted therapies modulate phosphoproteomic states, which are likely better predictors of drug responses than DNA or mRNA alterations. For example, increased EGFR/HER2 activating phosphorylation levels is a predictive marker for the EGFR/HER2 inhibitors. Nearly half of ERBB2/EGFR-mutated tumors, however, do not lead to high levels of EGFR/HER2 activating phosphorylation as confirmed by our calculations. By assessing multiple data modalities at once, we expect tools such as REFLECT will improve the fraction of benefiting patients while providing more reliable target matching in precision oncology. Indeed, the validations of REFLECT showed that targeting recurrent DNA, mRNA, and protein co-alterations all have beneficial outcomes, suggesting all data modalities add value. There is, however, a great need for improvement in the accuracy and completeness of the precision oncology knowledge bases to leverage multi-omic data for therapy selection. We anticipate that our knowledge on how to target mRNA and protein expression alterations will improve significantly in the near future. There is also a need for a better assessment of drug combination toxicities through the use of emerging informatics resources. Therefore, it is imperative that strategies such as REFLECT remain agile and adapt to evolving multi-omic data, precision oncology resources, and the clinical landscape to improve treatment selection for cancer patients.

With expanding data from precision therapy trials and genomics studies, it will become increasingly feasible to match combination therapies to complex molecular signatures. As lack of durable responses to genomically matched monotherapies significantly limits the population of benefiting cancer patients, REFLECT is potentially a critical addition to the current precision medicine paradigm. Our analyses suggest > 40% of patients may be potentially eligible for precision combination therapies when multi-omic data is utilized to identify co-actionable alterations. We expect this eligibility rate will be reduced due to other critical considerations such as drug toxicity. Nevertheless, this already corresponds to a substantial improvement over the conventional paradigm that relies on only DNA alterations as recent estimates suggest 15% of all cancer patients are eligible for precision therapies⁷². Therefore, we expect REFLECT will be a commonly used tool and resource to select precision combination therapies in both preclinical and clinical studies. We conclude a revolution in precision medicine to significantly improve benefiting patient populations and response durations is possible through clinical bioinformatic analyses of multiple data types with predictive tools such as REFLECT.

Materials and Methods

Multi-omic data

We used phospho-proteomics datasets of TCGA patients (TCPA) and cancer cell lines (MCLP) measured by reverse-phase protein array (RPPA), which were available from The Cancer Proteome Atlas Portal (https://tcpaportal.org/). Level 4 data (batch normalized across disease types) were downloaded for both patients and cell lines. To remove missing values, we filtered both data to maximize the number of samples and the number of proteins with no missing values. As a result, the patient dataset had 7544 samples by 179 (phospho-)proteins from 32 cancer types (no RPPA data was available for the LAML cancer type), and the cell line dataset had 397 samples by 179 (phospho-)proteins from 18 lineages. Because of their overlapping coverages of diverse disease types, we assumed that the patient dataset and the cell line dataset follow the same distribution of protein expressions across samples. To remove the batch effects and merge the two datasets, we standardized each dataset using Z-score transformation so that each protein had a zero mean and a unit variance. Then, the two standardized datasets were combined to form an integrated phospho-proteomics dataset.

For mRNA expression, we used RNA-seq datasets of TCGA patients (N = 8,887) and CCLE cell lines (N = 1,156). The batch effects within each dataset from diverse cancer types had already been corrected for and validated using the EB++²⁸. Using the same integration strategy for the phospho-proteomics data, we Z-score standardized the log2-transformed mRNA expressions in each RNA-seq dataset and then combined the two datasets together. To focus on essential genes relevant to cancer, in the current implementation of REFLECT features, we used genes from the COSMIC CGC³¹ v86. The oncogenic and tumor suppressor roles of genes are assigned based on the COSMIC CGC annotations. For actionable target selection, we used ∣Z-score∣ > 2 as a threshold to detect protein and mRNA alterations and match to therapy.

For genomic alterations (mutations and copy number alterations), we used patient datasets from TCGA (N = 8,764) and cell line datasets from the COSMIC-CLP (N = 996). The cancer-critical genes from the COSMIC CGC v73 were used for REFLECT calculations. For the mutation data, we filtered out hypermutated samples. Then, mutations were discretized to binary values (mutant = 1, wild type = 0). The copy number alterations were discretized to {−1, 0, 1}, representing deletion, wild type, and amplification, respectively. To account for differences in exome design, depth coverage, and other batch effects, we used the multi-site, multi-disease, consensus pan-cancer MAF file generated by the TCGA Multi-Center Mutation Calling in Multiple Cancers (MC3) project²⁹. For copy number analysis, we used the significantly altered copy number events based on GISTIC2.0 analysis of pan-cancer data (Affymetrix SNP 6.0) that had passed through the standardized TCGA data processing pipeline for all cancer types.

Genomic stratification markers

We selected actionable genomic alterations as the stratification markers from three sources. First, we selected 15 genomic vulnerability markers from NCI-MATCH (Molecular Analysis for Therapy Choice)³⁰. Then, from potentially actionable alterations of the MD Anderson Precision Oncology Knowledge Base²⁰, we extracted 182 genomic alteration markers that were not overlapped with the NCI-MATCH markers. As a result, these markers also covered the actionable genes listed in OncoKB. For each marker, the alterations included mutations, gene amplification/deletions, and specific amino acid substitutions (e.g., BRAF V600 mutation). Depending on specific genes, activating mutations were combined with amplification, and inactivating mutations were combined with gene deletions. Finally, we constructed a cohort of patients carrying potential immunotherapy response markers and 3 other cohorts based on the existence of key oncogenic alterations - MYC amplifications, TP53 hotspot mutations, and KRAS mutations.

We termed these markers as “master” biomarkers (N = 201). Annotations based on Gene Ontology (GO) biological process terms were used to categorize the master biomarkers. The top three GO terms were used; terms were ranked by evidence types and counts as provided by the GO database. Code is provided at https://github.com/cannin/extract_go_slim_annotation.

Sparse hierarchical clustering

Sparse hierarchical clustering identifies clusters driven by the most discriminant features³². The input to the algorithm is X, an n × p data matrix with n samples (e.g., patients) by p features (e.g., genes). Let x_i be a sample that is a p-dimensional vector indexed by i, where 1 ≤ i ≤ n, and d(x_i, x_i′) be a measure of the distance between samples x_i and x_i′. We assume that the distance measure is additive in features; that is, $d({\mathbf{x}}_i,{\mathbf{x}}_{i^{\prime }})={\sum }_j^p{d}_{i,i^{\prime },j}$, where d_i,i′,j is the distance measure between samples x_i and x_i′ along feature j.

The distance is defined as either the squared Euclidean distance d_i,i′,j = (X_ij – X^i′j)² or the absolute difference d_i,i′,j = ∣X_ij – X_i′j∣. The squared Euclidean distance, due to the square operation, assigns a relatively higher weight to larger values, while the absolute difference treats all values uniformly. In REFLECT, to smoothen small differences and strengthen big differences, the squared Euclidean distance was employed as the distance measure for continuous data types (i.e., protein and mRNA expressions). For discrete data types (i.e., mutations and copy number alterations), where only a small finite set of values exists, that is, {−1, 0, 1}, we used the absolute difference as their distance measures.

The sparse hierarchical clustering can be derived from the standard hierarchical clustering formulation. Note that the standard hierarchical clustering solves a constrained maximization problem

$$\underset{\mathbf{U}}{\text{maximize}}\left\{\sum _j\sum _{i,i^{\prime }}{d}_{i,i^{\prime },j}{U}_{i,i^{\prime }}\right\},$$

$$\text{subject to}\sum _{i,i^{\prime }}{U}_{i,i^{\prime }}^2\le 1,$$

which leads to the optimal solution $U_{i,i^{\prime }}^{\ast }\propto {\sum }_j{d}_{i,i^{\prime },j}$. The sparse hierarchical clustering can be formulated as the solution to a weighted constrained optimization problem

$$\underset{\mathbf{w},\mathbf{U}}{\text{maximize}}\left\{\sum _j{w}_j\sum _{i,i^{\prime }}{d}_{i,i^{\prime },j}{U}_{i,i^{\prime }}\right\},$$

$$\text{subject to}\sum _{i,i^{\prime }}{U}_{i,i^{\prime }}^2\le 1,\Vert \mathbf{w}{\Vert }^2\le 1,\Vert \mathbf{w}{\Vert }_1\le s,w_j\ge 0\forall j,$$

where w_j is a weight for feature j, and s is a tuning parameter that controls sparsity of w. Let U** optimize the above criterion, from which we have $U_{i,i^{\prime }}^{\ast \ast }\propto {\sum }_j{w}_j{d}_{i,i^{\prime },j}$, a weighted distance with feature j having a weight w_j. Due to the constraints, a small value of s makes w sparse, so U** depends only on a small subset of features. The optimization is iterated until convergence (w is fixed while U is optimized, and next U is fixed while w is optimized). After it is converged, hierarchical clustering is performed on the optimal solution U**. The output is hierarchical clusters of samples defined by a small set of discriminant features corresponding to nonzero elements of the converged w. In REFLECT, we employed the sparcl R package (implemented by Witten and Tibshirani)³² to execute the sparse hierarchical clustering.

Concordance of clusters with cancer types

To evaluate the association of cancer types with the clustered hierarchy of samples, represented as a dendrogram, we calculated the concordance of clusters with cancer types as

$$\text{Concordance}=\frac{{\sum }_{i=2}^n\mathbb{I}(L_i=L_{i-1})}{n-C},$$

where i denotes indices of samples along the clustered dendrogram, L_i is the label of the cancer type to which the sample i belongs, n is the number of samples, and C is the number of cancer types. The denominator n – C normalize the range of the concordance to be within 0 and 1. A large concordance implies a strong association between clustering and cancer types. When successive clustered samples always belong to different cancer types, the concordance becomes 0; when all samples of the same cancer types are tightly connected, the concordance is 1.

Analysis of tumor proliferation scores and REFLECT signatures

To evaluate the association of proliferation with the clustered hierarchy of samples, represented as a dendrogram, we calculated the Spearman’s correlation of sample rank along the dendrogram with proliferation scores. To calculate proliferation scores, we obtained marker genes of proliferation from Ref.³³, which includes MYBL2, BUB1, PLK1, CCNE1, CCND1, CCNB1, AURKA, E2F1, FOXM1, MKI67, MCM2, MCM3, MCM4, MCM5, and MCM6. Using log2-transformed RNA-seq data of TCGA samples, the proliferation score of a sample was calculated as the mean of expressions of its proliferation genes.

Cell lines and reagents

The cell lines used in experiments included MFE-296, HEC50B, DU4475, BT-474, HCC1954, MDA-MB-468, SK-OV-3, KM12, HCC2998, SW480, HCT15. They were obtained through the MD Anderson Characterized Core Cell Line Facility and the MDACC CRC Moon Shot program. Cell lines were authenticated by fingerprinting using short tandem repeat testing. The absence of mycoplasma contamination was also verified. Cells were maintained in RPMI-1640 supplemented with 10% heat-inactivated FBS and 1% penicillin with streptomycin. The targeted compounds (PARP inhibitor – Olaparib, Chk2 inhibitor – BML-277, AKT inhibitor MK-2206, MET inhibitor – JNJ-38877605, HER2 inhibitor – Neratinib, AURKA inhibitor – Alisertib, CDK4/6 inhibitor – Ribociclib, MAP3K4 inhibitor – Doramapimod (directly inhibiting downstream signaling partner, p38/MAPK14), PIK3CA inhibitor – Buparlisib) were obtained from Selleckchem.

Statement of Significance.

We developed the predictive bioinformatics platform, REFLECT, and a multi-omics based precision combination therapy resource. The REFLECT-selected therapies lead to significant improvements in efficacy and patient survival in pre-clinical and clinical settings. Use of REFLECT can optimize therapeutic benefit through selection of drug combinations tailored to molecular signatures of tumors.