Table 1.
Machine learning algorithm predicts PD-L1, TMB, TME in lung cancer
| Omics | Category | Task | Secondary task | Algorithm | Year | Description |
|---|---|---|---|---|---|---|
| Radiomics | PD-L1 | Expression | Prognosis | RF | 2020 [22] | Extracting image features from CT images to predict PD-L1 expression level and progression risk |
| Radiomics | PD-L1 | Expression | SResCNN | 2021 [14] | Using SResCNN to analyze PET/CT images and clinical data, using DLS score to predict PD-L1 expression | |
| Radiomics | PD-L1 | Expression | Logistic regression, RF | 2020 [25] | Extracting features from CT, PET, and PET/CT images to model and predict the positive and high expression of PD-L1 simultaneously | |
| Radiomics | PD-L1 | Expression | Survive | DL | 2020 [5] | Using deep learning to find CT image features to distinguish TMB expression and to predict survival in patients treated with ICIs |
| Pathomics | PD-L1/TMB | Expression | Treatment | ML | 2023 [29] | Extraction of the tumor, mesenchymal, and TIL counts from HE-stained images for modeling and assessment of TMB and PD-L1 expression levels and efficacy prediction |
| Multi-omics | PD-L1 | Treatment | ML | 2022[30] | Combining sequencing data, IHC images, demographic data and laboratory data to predict the efficacy of immunotherapy | |
| Multi-omics | PD-L1 | Expression | Pneumonia | LCI-RPV | 2023 [20] | The LCI-RPV model was developed to predict the ratio of PD-L1 expression to pneumonia by collecting CT images, CD274 counts and PD-L1 mRNA expression data |
| Multi-omics | TMB | Expression | ML | 2022 [31] | Combining genomic and epigenetic data to predict TMB | |
| Radiomics | TME | Prognosis | Treatment | ML | 2020 [39] | Extracting PET/CT image features to + distinguish groups who benefit from immunotherapy |
| Radiomics | TME | Expression | Treatment | ML | 2022 [37] | Predicting TME by modeling PET/CT image features with CD8+T expression data to predict the immune status |
| Radiomics | TME | Expression | Prognosis | ML | 2022 [38] | Extracting pGGO features from CT images combined with associated risk genes modelling to predict TME |
| Pathomics | TME/TIL | Expression | Prognosis | CNN | 2018 [40] | Use CNN to analyze HE images in the database, model and predict TME and OS |
| Pathomics | TIL | Expression | Prognosis | CNN | 2022 [41] | Development of I-score to predict clinical risk using CNN analysis of CD3+ T cell and CD8+T cell densities in WSI images |
| Pathomics | TME | Prognosis | CNN | 2020 [42] | Improved boundary recognition for WSI images, extraction of spatial features modeling prognosis | |
| Pathomics | TIL | Prognosis | Lunit SCOPE IO | 2022 [43] | Segmentation and quantification of WSI images to build the model Lunit SCOPE IO analysis TIL | |
| Multi-omics | TIL | Prognosis | Unsupervised clustering | 2022 [44] | Extraction of TIME, patient survival data, SMG and CNV modeling to analyze TIL | |
| Multi-omics | TME | Prognosis | Treatment | ML | 2022 [45] | Screening gene combinations and modelling to predict OS and efficacy |
| Multi-omics | TME | Expression | K-means, SVM | 2022 [46] | Screening, modeling, and predicting TIME of gene profiles using K-means and SVM |
PD-L1 Programmed Death Ligand 1, TMB Tumor Mutation Burden, TME Tumor Microenvironment, CT Computer Tomography, RF Random Forests, SResCNN Small Residual Product Network, LightGBM Light Gradient Boosting Machine, DL Deep Learning, ML Machine Learning, ICI Immune Checkpoint Inhibitor, WSI Whole Slide Image, TIL Tumor Infiltrating Lymphocyte, CNN Convolutional Neural Networks, SVM Support Vector Machine, SMG Significantly Mutated Gene, CNV Copy Number Variation, TIME Tumor Immune Microenvironment, OS Overall Survival, pGGO Pure Ground-Glass Opacity