A Collaborative Approach to Improving Missense Mutational Effect Predictions in Oncoproteins
Tekpinar, A. D.; Tekpinar, M.
Show abstract
Missense mutations in oncoproteins play a key role in cancer, which makes their accurate classification essential for cancer diagnosis and treatment. Although various variant effect prediction algorithms have been developed, their accuracy varies across different proteins. In this study, we introduce the APE score, a novel ensemble metric that integrates three state-of-the-art unsupervised predictors--AlphaMissense, PRESCOTT, and ESM1b. We evaluated APE scores performance using nine deep mutational scanning (DMS) experiments on five oncoproteins and a dataset of 1068 clinically labeled variants from 24 oncoproteins. Our results show that APE score is better than its individual components in six out of nine DMS experiments and achieves superior accuracy in distinguishing pathogenic variants from benign ones. Additionally, we established classification thresholds to categorize 444087 variants into likely pathogenic, variants of unknown significance (VUS), and likely benign categories. The APE score reduces misclassification rates and provides a more reliable tool for assessing the functional impact of missense mutations in cancer-related proteins. All scores and classifications are provided in a publicly available repository to support further research and clinical applications.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- 3Cnet: Pathogenicity prediction of human variants using knowledge transfer with deep recurrent neural networks 96%
- MutaFrame - an interpretative visualization framework for deleteriousness prediction of missense variants in the human exome 95%
- Combining protein sequences and structures with transformers and equivariant graph neural networks to predict protein function 95%
Similar papers in this journal
- Cancer SIGVAR: A semi-automated interpretation tool for germline variants of hereditary cancer-related genes 96%
- KATK: fast genotyping of rare variants directly from unmapped sequencing reads 93%
- dbMTS: a comprehensive database of putative human microRNA target site SNVs and their functional predictions 92%
Similar papers in this journal
- TripletProt: Deep Representation Learning of Proteins based on Siamese Networks 94%
- Rapid Reconstruction of Time-varying Gene Regulatory Networks with Limited Main Memory 92%
- CPGL: Prediction of compound-protein interaction by integrating graph attention network with long short-term memory neural network 92%
Similar papers in this journal
- Novel ratio-metric features enable the identification of new driver genes across cancer types 94%
- Uncovering Hidden Cancer Self-Dependencies through Analysis of shRNA-Level Dependency Scores 94%
- Protein model accuracy estimation empowered by deep learning and inter-residue distance prediction in CASP14 94%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.