nicolaengelbrecht/pathogenicity-predictors-honours

Code and scripts for the Honours project “Pathogenicity predictions in less-common breast-cancer genes”. This repository contains data preprocessing steps, statistical analyses, and figure generation workflows used to investigate the performance of 53 pathogenicity prediction tools on variants derived from African breast cancer patient data. # pathogenicity-predictors-honours Code and scripts for the Honours project “Pathogenicity predictions in less-common breast-cancer genes”. This repository contains data preprocessing steps, statistical analyses, and figure generation workflows used to investigate the performance of 53 pathogenicity prediction tools on variants derived from African breast cancer patient data. Python was used for all preprocessing steps, including merging the VEP and VCF files, filtering for canonical transcripts, removing population polymorphisms, excluding haplotype calls (0/0), and eliminating duplicate rows. R was used for downstream statistical analyses and visualisations. This included dimensionality reduction (PCA, MDS, t-SNE), group comparisons (ANOVA), and clustering approaches such as dendrograms and heatmaps. Additional visualisations included scatter plots and agreement matrices. The design of the mathematical model was also implemented in R, where predictor outcomes were integrated into a conceptual logistic framework to demonstrate how multiple tools can be combined into a single model.