In the article entitled, “A New Approach to Drug Repurposing with Two-Stage Prediction, Machine Learning, and Unsupervised Clustering of Gene Expression” (OMICS: A Journal of Integrative Biology: 2022;26(6):339–347) by Cong et al., the authors correct and update their original analysis in which they overlooked a GEO2R tool update (GEO2R, 2020). The authors apologize for this error.
In the corrected article, the authors update the related text, Figures 2 and 3, Table 2, and Supplementary Tables S1–S9 and S13.
FIG. 2.
Discovery of differentially expressed genes. (a) The differentially expressed genes for IBM. (b) The differentially expressed genes for PM. (c) The differentially expressed genes for DM. The vertical axis reflects the intentionality of the statistic (−log10[p-value]) and the horizontal axis reflects the magnitude of the change (log2[fold-change]). Red and blue colors indicate positive and negative directions of significant expression changes, and gray indicates changes below the significance level. DM, dermatomyositis; IBM, inclusion body myositis; PM, polymyositis.
FIG. 3.
Number of small-molecule compounds for each disease. Numbers in overlapping regions are compounds shared between the disorders.
Table 2.
Breakdown of Common Drugs
|
The new method developed and reported in the authors' original article remains the same and requires no corrections.
The corrected online article and the Supplementary Tables are available at: https://www.liebertpub.com/doi/10.1089/omi.2022.0026
The Abstract originally read:
“…Consequently, we found disease-specific gene expression, and importantly, identified 20 drugs such as BMS-387032, phorbol-12-myristate-13-acetate, mitoxantrone, alvocidib, and vorinostat as candidates for repurposing…. That is, inclusion body myositis and DM. The two-stage prediction approach to drug repurposing….”
The updated Abstract now reads:
“….Consequently, we found disease-specific gene expression, and importantly, identified 22 drugs such as KM 00927, I-BET, alvocidib, and vorinostat as candidates for re-purposing…. The two-stage prediction approach to drug repurposing….”
On page 344, the 1st paragraph originally ended:
“….(Supplementary Tables S3—S5).”
The updated ending of the 1st paragraph now reads:
“….(Supplementary Tables S3 and S4).”
On page 344, the 2nd paragraph originally ended:
“.…are listed in Supplementary Table S6.”
The updated ending of the 2nd paragraph now reads:
“….are listed in Supplementary Table S5.”
On page 344, the 3rd paragraph originally ended:
“….and Supplementary Table S7). We found 10 drugs that corresponded to IBM and PM, 10 drugs that corresponded to PM and DM, and 14 drugs that corresponded to all three diseases.”
The updated ending of the 3rd paragraph now reads:
“….and Supplementary Table S6). We found nine drugs that corresponded to IBM and PM, four drugs that corresponded to IBM and DM, nine drugs that corresponded to PM and DM, and ten drugs that corresponded to all three diseases.”
On page 344, the 4th paragraph originally ended:
“….(Supplementary Tables S8 and S9). The following L1000 perturbation gene signatures scatter plots were generated:
IBM: https/maayanlab.cloud/clustergrammer/11000cds2/61082faed99ec600506a634e
PM: https/maayanlab.cloud/clustergrammer/11000cds2/612306c0d99ec600506a6a88
DM: https.maayanlab.cloud/clustergrammer/11000.cds2/6123071fd99ec600506a6a8a”
The updated ending of the 4th paragraph now reads:
“….(Supplementary Tables S7 —S9). The following L1000 perturbation gene signatures scatter plots were generated:
IBM: https://maayanlab.cloud/clustergrammer/11000cds.2/63c5271f7903f4005f0b012a
PM: https://maayanlab.cloud/clustergrammer/11000cds.2/63c5285e7903f4005f0b012c
DM: https://maayanlab.cloud/clustergrammer/11000cds.2/63c52f867903f4005f0b0134”
On page 345, beginning near the end of the 1st full paragraph, the text originally read:
“….based on the same original data. This yielded 20 small-molecule compounds for only one disease, for which applicability to other diseases was unconfirmed, 10 with unconfirmed efficacy only for DM, and 10 with unconfirmed efficacy only for IBM.
Of the 421 genes directly affected by the 10 compounds that were validated for IBM and PM using L1000CDS2 but not validated for DM, 224 were common to the disease-specific gene expression data for DM obtained using GEO2R. Of the 407 genes directly affected by the 10 compounds whose efficacy was unconfirmed only for IBM, 307 were common to the disease-specific genes for IBM obtained using GEO2R.
Next, we obtained BRD-IDs and PubChem Names for 20 small-molecule compounds with unconfirmed efficacy for only one of the three diseases using side effect prediction based on the L1000 data (http://maayanlab.net/SEP-L1000/index.html). PubChem names were obtained and their efficacy as drugs was investigated in several databases (Supplementary Table S13). Of the 10 compounds that were only investigated for IBM and PM, AT-7519, CGP-60474, BMS-387032, and alvocidib inhibit cyclin-dependent kinases; mitoxathrone inhibits DNA cleavage by topoisomerase II; dFdCTP, a conversion product of gemcitabine, competes with deoxycitine tri-phosphate (dCTP) to inhibit DNA replication (Chabner et al., 2011); SB-218078 inhibits check-point kinase 1 (Chk1); and CHR-2797 converted into a pharmacologically active acid product named CHR-79888 inside cells that inhibits the M1 family aminopeptidases; all of these com-pounds ultimately induce apoptosis.
Wortmannin is a cell-permeable Phosphatidylinositol-3-kinase (PI3K) inhibitor. Phorbol-12-myristate-13-acetate is a potent tumor promoter that is often used in biomedical research to activate the signal transduction enzyme protein kinase C, and is linked to gene transcription, cell proliferation, differentiation, programmed cell death, and immune pathways.
Of the 10 compounds that were only validated for PM and DM, BRD-K19181733 is a receptor antagonist that inhibits and attenuates dopaminergic effects. Perhexiline maleate is a potent inhibitor of carnitine palmitoyltransferase 1. Selumetinib and 7878890 are highly selective methyl ethyl ketone inhibitors. WZ-4-145 is a selective inhibitor of epidermal growth factor re-ceptor tyrosine kinase activity, which inhibits cell growth and proliferation. Wiskostatin and BRD-A90643929 inhibit protein synthesis and function. RO-28-1675 is a potent allosteric gluco-kinase activator. CD-1530 markedly enhances catalytic activity, increases glucose metabolism, and lowers blood glucose levels. The mechanism of action of cyclosporin-a is not clear, but it is thought to bind to cytophilin and inhibit calcineurin.”
The updated text for the above is:
“….based on the same original data. This yielded 22 small-molecule compounds for only one disease for which applicability to other diseases was unconfirmed, nine with un-confirmed efficacy only for DM, four with unconfirmed efficacy only for PM, and nine with unconfirmed efficacy only for IBM.
Of the 266 genes directly affected by the nine compounds that were validated for IBM and PM using L1000CDS2 but not validated for DM, 112 were common to the disease-specific gene expression data for DM obtained using GEO2R. Of the 63 genes directly af-fected by the four compounds whose efficacy was unconfirmed only for PM, 53 were com-mon to the disease-specific genes for PM obtained using GEO2R. Of the 171 genes directly affected by the nine compounds whose efficacy was unconfirmed only for IBM, 123 were common to the disease-specific genes for IBM obtained using GEO2R.
Next, we obtained BRD-IDs and PubChen Names for 22 small-molecule compounds with unconfirmed efficacy for only one of the three diseases using side effect prediction based on the L1000 data (http://maayanlab.net/SEP-L1000/index.html). PubChem names were obtain and their efficacy as drugs was investigated in several databases (Supplemen-tary Table S13). Of the nine compounds that were only validated for IBM and PM, BRD-A76490030 is functionally related to certain α-amino acids, and is predicted to be an EGFR inhibitor. PARP inhibitor and calcium channel blocker at L1000 fireworks display (L1000FWD), although no specific functions have been reported, Brefeldin A is a fungal metabolite with broad antibiotic activity. Curcubitacin 1 is a plant metabolite and a cucur-bitacin that acts as an antitumor agent. I-BET151 induces apoptosis via down-regulation of BCL2 or induces G0/G1 cell cycle arrest. Narciclasine has been reported to exhibit signifi-cant anticancer activity in experimental models of brain tumors by targeting GTPases and damaging actin cytoskeletal organization (Van Goietsenoven et al., 2013). Radicicol is a macrocyclic antifungal antibiotic that inhibits signal-dependent transcriptional activation and induces differentiation in leukemia. TG101348 is an ATP-competitive inhibitor of Janus-associated kinase 2 (JAK2) and FMS-like tyrosine kinase 3 (FLT3; CD135; STK1; FLK2) used in the treatment of primary and secondary myelofibrosis. Tyrphostin AG 1478 may have a role as an epidermal growth factor receptor antagonist, antineoplastic agent, geroprotector, and antiviral agent. Vorinostat is a histone deacetylase (HDAC) inhibitor used to treat cutaneous symptoms in patients with cutaneous T-cell lymphoma (CTCL) after systemic therapy.
Of the four compounds that were only validated for IBM and DM, only PK-11195 has been found to act in the assessment of neuronal damage (Weissman and Ravch, 2003). Although the pharmacological actions of the other three have not been reported, there is a clear trend in MOA prediction in the L1000FWD database: BRD-A28970875 has a proba-bility of 0.9477, and SARMENTOGENIN has a probability of 0.4499, that is a protein synthesis inhibitor. The last one, KM 00927, has a 0.6189 chance of being an NFkB path-way inhibitor.
Of the nine compounds that were only validated for PM and DM, A-443654 is a pan-Akt inhibitor with equivalent activity against intracellular Akt1, Akt2 or Akt3, in-hibiting tumor cell growth and proliferation. Alvocidib as an inhibitor of cyclin-dependent kinase (CDK), inhibits CDK phosphorylation, down-regulates cyclin D1 and D3 expression, and induces G1 cell cycle arrest and apoptosis. BRD-K41859756 aka NVP-AUY922 is a po-tent HSP90 inhibitor that effectively destabilizes IGF-1Rβ protein and is involved in cell growth inhibition, autophagy and apoptosis. CGP-60474 is one potent anti-endotoxin agent, effective in inhibition against CDK. EI-346 has a predicted probability of 0.2968 as a cyclo-oxygenase inhibitor at L1000FWD. I-BET is considered a BET bromodomain inhibitor and is described in the related link for I-BET 151 above. NVP-AUY922 can effectively down-regulate and destabilize IGF-1Rβ protein, leading to growth inhibition, autophagy and apoptosis. PD-0325901 targets mitogen-activated protein kinase (MAPK/ERK kinase or MEK) and is expected to have antitumor activity. QL-X-138 is a potent and selective BTK/MNK dual kinase inhibitor that may be useful in the study of B-cell malignancies.
On page 345, beginning near the end of the 3rd paragraph up from end of the page, the text originally read:
“….related to immune responses. Since IBM and DM are both autoimmune and inflam-matory muscle diseases (Nishino, 2020), it is likely that the reactions inhibited by the 10 vali-dated compounds need to be suppressed for treatment.
The two-stage prediction approach to drug repurposing presented here offers innovation to inform future drug discovery and clinical trials in a variety of human diseases. We predict that drugs shown to be effective for only two diseases may also be effective for other diseases, IBM and DM, for which no efficacy was reported previously.”
The updated text for the above on page 345 is:
“….related to immune responses. Since these three diseases are both autoimmune inflammatory muscle diseases (Nishino, 2020), it is likely that the reactions inhibited need to be suppressed for treatment.
The two-stage prediction approach…may also be effective for other diseases, for which no efficacy was reported previously.”
On page 346, the 1st reference in the “References” has been removed from the original article:
“Chabner BA, Amrein PC, Druker BJ, et al. (2011), Antineoplastic agents. In: Goodman & Gilman's The Pharmacological Basis of Therapeutics, 12th ed. Brunton LL and Parker KL, eds. New York: McGraw-Hill Education Press, 1315—1403.”
On page 347 in the “Abbreviations Used,” the following 3 abbreviations have been removed from the original article:
“Chk1 = checkpoint kinase 1
dCTP = deoxycytidine triphosphate
PI3K = permeable Phosphatidylinositol −3-kinase”
The updated Figures 2 and 3 (with no changes in legends) and updated Table 2 (with updated footnotes) appearing in the updated article on pages 344 and 345 are:
Address correspondence to:
Toshinori Endo, PhD
Laboratory of Information Biology
Information Science and Technology
Hokkaido University
Sapporo 060-0814
Japan
E-mail: endo@ist.hokudai.ac.jp
Reference
- Cong Y, Shintani M, Imanari F, et al. A new approach to drug repurposing with two-stage prediction, machine learning, and unsupervised clustering of gene expression. OMICS 2022;26(6):339–347; doi: 10.1089/omi.2022.0026 [DOI] [PMC free article] [PubMed] [Google Scholar]