Comprehensive assessment of computational algorithms in predicting cancer driver mutations

What this source contributed

Records in this release that cite this source. A record can cite several sources.

• 28 methods
• 33 configurations
• 4 protocols
• 2 datasets
• 100 evaluations
• 500 results
• 42 evidence claims
• 1 use case

Results

Values as the source printed them. Compare results only under compatible protocols, datasets and metrics.

Claims

Statements this source supports, each with where in the source it was checked.

• Baseline: Assess somatic small-variant oncogenicity: Baseline ClinGen/CGC/VICC rules (OncoVI) and SIFT (uc-clinical-20260930-source-oncovi: Methods (IF-ELSE implementation of the points-based SOP); somatic-oncogenicity-20261009-source-chen2020: Table 1 (SIFT position-specific scoring matrix) and Introduction (SIFT widely used in cancer research))
• CADD: description: Description Combined Annotation-Dependent Depletion: a genome-wide variant deleteriousness score that integrates more than 60 genomic features in a machine learning model trained to separate simulated de novo variants from variants fixed in human populations since the human-chimpanzee split. It scores single nucleotide variants and short insertions and deletions anywhere in the reference assembly, not only missense variants. (Record description; cited Rentzsch et al. 2019, Nucleic Acids Research 47:D886)
• CanDrA: description: Description Scores missense variants with a support vector machine over structural, evolutionary and genomic features together with the scores of published predictors. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• CHASM: description: Description Scores missense variants with a random forest over structural, evolutionary and genomic features. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• CTAT (cancer and population ensembles): description: Description Scores missense variants with principal-component ensembles of other predictors: the cancer ensemble combines TransFIC, FATHMM, CHASM and CanDrA, and the population ensemble combines SIFT, PolyPhen-2, MutationAssessor and VEST. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• DANN: description: Description Scores variants with a deep neural network over conservation, epigenetic signals, the scores of other predictors and genetic context. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• DEOGEN2: description: Description Scores missense variants with a random forest over evolutionary, protein, gene and pathway features together with PROVEAN scores. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• Eigen: description: Description Scores variants by an unsupervised spectral method that combines the scores of other tools, allele frequencies and epigenomic signals; the paper evaluates both the Eigen and the Eigen-PC variant. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• Evidence summary: somatic small-variant oncogenicity: Summary Two predictor comparisons join the existing OncoVI evidence; their values cannot be pooled. In Chen et al. 2020 (peer reviewed), 33 predictors scored somatic missense mutations labelled by OncoKB curation or by the authors' own cell viability assays, with likely neutral or assay-neutral somatic negatives. The authors co-developed one scored tool, CanDrA, and developed none of the others. With each tool's own default calls, the highest accuracy was 0.70 on OncoKB labels (DEOGEN2, sensitivity 0.87, specificity 0.51), and several tools called most neutral mutations damaging, for example M-CAP with sensitivity 0.97 and specificity 0.11, and CanDrA with specificity 0.17 and the lowest accuracy, 0.49. On the assay labels the highest default accuracy was 0.65 (PrimateAI). When each tool's median score on the test draw was used as the threshold, which ranks tools but is not an operating point a reviewer could apply, the highest accuracy was 0.69 on OncoKB labels (PROVEAN and VEST4) and 0.70 on assay labels (CTAT-cancer and PrimateAI), and FATHMM-disease scored 0.50 on OncoKB labels. Chen et al. print each value as a mean with a range of two standard deviations over 100 draws, not a confidence interval, and do not say which OncoKB positive set fed these tables. In a single-author bioRxiv preprint that is not peer reviewed (Lee 2026), 49 scores were compared on Cancer Gene Census missense variants whose negatives are germline-benign labels (ClinVar benign or likely benign, and gnomAD common variants), which this use case does not accept as somatic controls. In that table 42 of the 49 scores had a lower AUROC on oncogene variants than on tumour-suppressor variants, for example AlphaMissense 0.885 against 0.900 and CADD 0.821 against 0.894, while popEVE went the other way (0.900 against 0.842). The highest all-gene AUROC was 0.876 (MutScore); these AUROC values are rounded here to three decimals from the table's full precision. None of these comparisons covers small indels, calibrated probabilities, criterion-level evidence strength, drug response or reviewer agreement. (somatic-oncogenicity-20261009-source-chen2020-additional-file-9 and -10: Benchmark 2; somatic-oncogenicity-20261009-source-chen2020-additional-file-21 and -22: Benchmark 5; somatic-oncogenicity-20261009-source-chen2020: Methods and reference 7; somatic-oncogenicity-20261009-source-lee2026-oncocal-tool-performance: tables/tool_performance.tsv; somatic-oncogenicity-20261009-source-lee2026: Methods and Results)
• FATHMM-MKL: description: Description Scores variants by multiple kernel learning over conservation and epigenomic signals. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• FATHMM-XF: description: Description Scores variants by multiple kernel learning over conservation, genomic features and epigenomic signals. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• FATHMM: description: Description Scores missense variants with hidden Markov models over sequence homology; the paper evaluates both the disease-weighted and the cancer-weighted variant. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• GenoCanyon: description: Description Scores a position by the posterior probability that it is functional, from unsupervised statistical learning over conservation and biochemical annotation. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• Integrated_fitCons: description: Description Scores the fitness consequences of a position by applying the INSIGHT evolutionary model to integrated epigenomic signals. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• LRT: description: Description Scores a variant by a likelihood ratio test of codon neutrality from sequence homology. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• M-CAP: description: Description Scores missense variants with a gradient-boosted tree classifier over the scores of published predictors and conservation. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• MetaLR: description: Description Scores missense variants with a logistic regression over nine prediction scores and 1000 Genomes allele frequencies. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• MetaSVM: description: Description Scores missense variants with a radial-kernel support vector machine over nine prediction scores and 1000 Genomes allele frequencies. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• MPC: description: Description Scores missense variants with a logistic regression over regional missense constraint, missense badness and PolyPhen-2 scores. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• MutationAssessor: description: Description Scores a substitution from sequence homology with a combinatorial entropy formalism. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• MutationTaster: description: Description Scores variants with a naive Bayes classifier over conservation, genetic context and regulatory features. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• MutPred: description: Description Scores missense variants with a random forest over protein structural and functional properties, conservation and SIFT scores. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• MVP: description: Description Scores missense variants with a deep neural network over sequence and structural features, the scores of published predictors and conservation. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• PolyPhen-2: description: Description Scores missense variants with a naive Bayes classifier over eight sequence-based and three structure-based features; the paper evaluates both the HumDiv and the HumVar trained variant. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)
• PrimateAI: description: Description Scores missense variants with a deep residual neural network over sequence homology. (somatic-oncogenicity-20261009-source-chen2020: Table 1, Summary of 33 computational algorithms included in this study)

Models, benchmarks and other records

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.