Analyzing Genetic Networks Using Neural Networks
summary
Claims that machine learning can recover gene regulatory wiring and shows a runnable pipeline that actually predicts the PurR regulon (31 known targets) from an ~4,000‑gene RNA‑seq matrix. Uses RegulonDB for training labels, PCA for noise reduction, SMOTE oversampling, and compares classifiers including linear SVM, random forest and a small Keras network. Builds Colab/Binder notebooks, evaluates on a held‑out test set of ~35 genes and reports high accuracy (≈94%), with code and visualizations working locally.
AI-written summary · about
chapters (17)
- 0:00Genetic networks and ML intro
- 3:23Learning genetic network wiring
- 8:44Gene function and DNA complexity
- 13:25Transcription factors and regulation
- 16:12Purine metabolism in E. coli case
- 19:11Setting up the machine learning pipeline
- 21:53Extracting gene regulatory networks
- 27:19Network structure and hub genes
- 31:32Clustering gene networks
- 36:25Purine metabolism subnetwork
- 40:15Expression data prep for ML
- 43:50Expression dataset structure
- 51:56Gene correlations and PCA filtering
- 58:33Annotating train/test datasets
- 1:04:51Training ML models on genes
- 1:11:25Nonlinear EDA methods
- 1:18:26Summary and future
related talks
39mIdentifying Disease Mechanisms Using Gene Networks as Priors
38mInterpreting Deep Learning Models Of Functional Genomics Data To Decode Regulatory Sequence...
40mGraph neural networks for biomedical insights
33mNetwork Data Integration Enables Precision Medicine
35mFrom cells to drug responses - machine learning in cancer research
01h 06mIntroduction to Biological Network Analysis III: Identifying Network Modules
35mData Driven "Pathway" Analysis with ADAGE
43mPanel on Interpretability in the Biological Sciences