There is no required reading until Week 3. Lecture slides are on Canvas.
Embedded Files
Lecture Week Date Theme
0. Introduction - course introduction, ML principles, and pitfalls
0. Introduction - course introduction, ML principles, and pitfalls
1 1 1/6 Course introduction; ML challenges in biological problems
Please see lecture slides on Canvas
2 1 1/8 Probabilistic modeling & learning; mathematical foundations
3 2 1/13 Deep learning (autoencoders, MLPs, CNNs, RNNs, and transformers)
Optional readings
4 2 1/15 Model interpretation techniques from the explainable AI field
Optional readings
https://christophm.github.io/interpretable-ml-book/shap.html
Molnar: Interpretable ML Book Chapter 5.10 (Shapley values)
Molnar: Interpretable ML Book Chapter 5.10 (SHAP)
5 3 1/20 Model interpretation techniques from the explainable AI field & Gene expression basics
Optional readings
https://christophm.github.io/interpretable-ml-book/shap.html
Molnar: Interpretable ML Book Chapter 5.10 (Shapley values)
Molnar: Interpretable ML Book Chapter 5.10 (SHAP)
6 3 1/22 Research example - From Model Explanations to Discovery: Explainable AI in Cancer Precision Medicine
1. Bulk Transcriptomics
1. Bulk Transcriptomics
7 4 1/27 L - Unsupervised learning of gene expression data
8 4 1/29 D - Linear models
L - Cancer biology
Required reading
Optional reading
"PAUSE: principled feature attribution for unsupervised gene expression analysis"
“Learning the parts of objects by non-negative matrix factorization”
Basic data interpretation techniques
9 5 2/3 Project Proposal Workshopping
D - Supervised transcriptome (Alzheimer's)
Required reading
“Unified AI framework to uncover deep interrelationships between gene expression and Alzheimer’s disease neuropathologies”
Optional readings
ML pitfalls
Schreiber et al. “A pitfall for ML methods aiming to predict across cell types”
DeGrave*, Janizek* and Lee. “AI for radiographic COVID-19 detection selects shortcuts over signal”
GSEA (classic: “On testing the significance of sets of genes”; recent: https://www.nature.com/articles/s41467-021-22862-1)
2. Single-cell Transcriptomics
2. Single-cell Transcriptomics
10 5 2/5 D - Supervised transcriptome (cancer)
L - Single-cell genomics
Required readings
Optional readings
“Biologically informed deep neural network for prostate cancer discovery”
“Predicting Drug Response and Synergy Using a Deep Learning Model of Human Cancer Cells”
11 6 2/10 D - Single-cell transcriptome (unsupervised)
L - Gene regulatory network
Required readings
Optional readings:
"MultiVI: deep generative model for the integration of multimodal data"
"Population-level integration of single-cell datasets enables multi-scale analysis across samples"
"Isolating salient variations of interest in single-cell data with contrastiveVI"
"PULSAR: a Foundation Model for Multi-scale and Multi-1 cellular Biology"
"Universal Cell Embeddings: A Foundation Model for Cell Biology"
3. Genetics/genomics
3. Genetics/genomics
12 6 2/12 D - Single-cell transcriptomics & gene regulatory network
L - Gene regulatory network
13 7 2/17 D - Single-cell foundation models
L - Genomics basics + Genome interpretation
14 7 2/19 D - Genomic profile prediction
L - Cell Painting
Required readings
Optional readings:
4. Emerging areas (cell painting & mRNA molecule design)
4. Emerging areas (cell painting & mRNA molecule design)
15 8 2/24 D - Genomic Interpretability
L - ML pitfalls
Required reading
Optional reading
16 8 2/26 D - Cell painting
Project Checkpoint Workshopping
5. Medicine & healthcare
5. Medicine & healthcare
17 9 3/3 D - Real-world clinical AI
L - Biological age
18 9 3/5 Guest Lecture - James Zou (Stanford)
19 10 3/10 D - Impact of aging on disease
D - Trustworthy & transparent medical AI
Required reading
Optional reading
Final
Final
20 10 3/12 Final project presentation
Report abuse