Using Support Vector Machines in Scikit Learn to discover genetic aetiologies
summary
Support vector machines implemented with scikit-learn + Pandas + Joblib process features from single SNPs up to functional gene‑sets and fail to improve classification AUC (≈0.6–0.7) over the conventional polygenic risk score, but do produce interpretable coefficients: repeated train/test splits show a consistent high weight on the FMRP gene set signal. The workflow uses randomized search CV + pipelines with scaling, explores ~5,000 sub‑threshold SNPs, and simulation experiments show kernels detect pairwise interactions that linear models miss.
AI-written summary · about
chapters (13)
- 0:00Schizophrenia genomics intro
- 1:55Genetic complexity explained
- 4:02Genome-wide association studies in psychiatry
- 7:13Polygenic scores vs ML
- 8:41Support Vector Machines fundamentals
- 10:44Linear SVM limits and kernel trick
- 12:10Feature selection for SVM
- 14:16Parameter tuning in scikit-learn
- 16:24Results and performance
- 19:57Gene set analysis with pandas
- 22:00Interpreting SVM coefficients
- 25:11Conclusions and future
- 30:48Q&A
related talks
29mDeveloping and Delivering Personalized Polygenic Scores at Scale
35mFrom cells to drug responses - machine learning in cancer research
32mHow machine learning helps cancer research
01h 17mMachine Learning with Scikit-Learn
30m23, Python and Me: Using Machine Learning in Python to Analyze Consumer Genomics Data
17mLeveraging Knowledge Graph and Imaging to Predict Gene–Disease Links in Cardiovascular Diseases
20mAdvancing Epigenetic Predictors with Scalable Machine Learning
42mPython’s role in accelerating biomedical science