How to Enroll
Recommendations: Python proficiency (e.g. CS 30 or CS 135 or DATA 201); basic molecular biology (e.g. BIO 13 plus (BIO 41 or BIO 105 or another BIO course numbered higher than 41)).
However, we may be able to help students lacking some background get up to speed on some background content, depending on student interests and needs.
Course Aims and Description
- comprehend and explain techniques underlying the modern AI methods used in biomedical research
- create, run, and edit code implementing such methods using Python and PyTorch
- understand and critique research papers in this exciting field
- propose and implement extensions of existing uses of AI in biology
- understand the limitations of such methods and implications for interpretation
These aims will be achieved through problem sets, reading and discussing literature, and a course project with multiple deadlines. There will be weekly assignments throughout the term.
Grading will be based on problem sets, literature presentations and discussions, and a team course project.
Course Requirements:
Prerequisites: Please fill out the course survey to describe your background in computer science, Python programming, and molecular biology.
Recommendations: Python proficiency (e.g. CS 30 or CS 135 or DATA 201); basic molecular biology (e.g. BIO 13 plus (BIO 41 or BIO 105 or another BIO course numbered higher than 41).
Readings:
We are considering two online textbooks as references; both are freely available in digital format and can be purchased as print copies. Deep Learning: Foundations and Concepts by Christopher M. Bishop with Hugh Bishop, published in 2024 by Springer Nature, offers a clear introduction to a wide range of modern methods. Deep Learning by Ian Goodfellow, Yoshua Bengio and Aaron Courville, published in 2016 by MIT Press, is a bit older but also more mathematical in nature.
In addition, students will read, present, and discuss a number of primary papers from the AI and biology literature. Some suggestions are listed below, though exact selections may vary.
Course structure:
In Fall 2026, the course will meet MW 1:30-2:45pm.
For the first part of the term, we will develop needed background in machine learning. Depending on the student population, we may also want to cover basic concepts in molecular biology describing key problems, including protein structure prediction, transcriptomics, gene reglation, and the drug discovery process. Homeworks using jupyter notebooks in Python will help ensure that students understand the concepts being discussed.
After the October holiday, we will transition to reading and discussing both ``classic'' (e.g. going back to 2014) and new (2026) research papers that implement some of these methods to solve biological problems. Students will both present papers to the class and discuss the content in groups. Towards the end of the semester we will include papers focusing on issues that can arise with AI in biology, building an understanding of some things that can go wrong.
Also in October, students will form project groups and begin project work rather than homework exercises. At the end of the term each student project team will present their work to the class. Projects may include evaluating or comparing existing methods, perhaps using new data, tuning models, and proposing improvements. Project suggestions will be offered, but focusing on your own interests is also encouraged.
Tentative Course Schedule:
(Please expect contents to change over time and in response to student interests; this is just a rough outline)| Week Starting | Mon | Weds |
| Sept. 7 | Course introduction, logistics | |
| Sept. 14 | Deep learning fundamentals | Model training and evaluation |
| Sept. 21 | Recurrent Neural Networks | Transformers, attention, LLMs |
| Sept. 28 | Convolutional neural networks | Autoencoders / VAEs |
| Oct. 5 | GNNs | diffusion / contrastive learning |
| Oct. 12 | NO CLASS, Tufts Holiday | Protein structure prediction: Alpha Fold (1-3) |
| Oct. 19 | Protein Language Models: ESM2 | Protein Language Models: protBERT |
| Oct. 26 | LLMs for biomedical text mining: bioBERT | LLMs for biomedical text mining: med-PALM |
| Nov. 2 | single cell models: scVI, scBERT | single cell models: Geneformer, scGPT |
| Nov. 9 | regulatory network inference: DeepSEM | Weds class on Tues: RegDiffusion |
| Nov. 16 | Drug repositioning: TxGNN | Training issues: fitting random data (Zhang+ ICLR 17) |
| Nov. 23 | Hallucination (Kalai and Vempala 24) | NO CLASS: Thanksgiving holiday |
| Nov. 30 | Bias, data leakage (ref TBA) | Benchmarking in the age of AI (Russell+, 26) |
| Dec. 7 | Agentic AI (Zou+ 25) | Presentations |
| Dec. 14 | Presentations | Reading period |