Goals

• Learn about machine learning methods applied to problems in structural biology
• Learn how to critically read and evaluate papers
• Learn how to pose research problems and practice oral and written scientific communication skills
• Bonus: Exposure to relevant basic and applied ML research in industry from guest speakers

Topics

A non-exhaustive list of topics we will cover include:

• An introduction to structural biology
• Protein structure prediction before and after AlphaFold2
• Computer vision and cryo-electron microscopy (cryo-EM)
• Computational protein design, in particular, antibody and vaccine design
• Physics-based modeling and statistical mechanics
• Small molecule drug discovery

Selected papers will cover a broad range of algorithmic concepts and machine learning techniques including:

• Supervised learning and designing appropriate benchmarks and metrics
• Language modeling and transformers
• Generative modeling techniques including VAEs, GANs, normalizing flows, and diffusion models
• Geometric deep learning
• Neural rendering and multi-view 3D reconstruction

In addition to the assigned papers, optional primers or reviews on relevant topics will be made available for background reading.

Assignments

Final project guidelines: link

Guest Speakers

Tuesday, March 3rd, 1:20pm ET
Mark Goldstein (Flatiron Institute)

Title: Diffusion Generative Models

Abstract: Generative models have become central tools in computational biology and beyond, enabling the design of proteins, antibodies, small molecules, and more. But what exactly is a generative model? In this lecture, we survey the landscape of deep generative modeling — or more plainly, ways to fit and sample from complex probability distributions using large neural networks.

We start with "early" approaches like normalizing flows and VAEs, make our way through energy-based models, and arrive at diffusion- and flow-based models. We discuss the intuition behind each paradigm, which limitations motivated the community to move on, and why diffusion/flow models have stuck around for some time. We also cover the basic design choices one faces when training or using a diffusion model in practice.

We then turn to applications in protein design, examining how these frameworks have been adapted to handle the geometric and biochemical structure of biomolecules. We will see a few successful applications, illustrating how the principles from the first half of the lecture come to life in state-of-the-art systems for controllable structure and sequence design.

Bio: Mark Goldstein is a Research Fellow in the Center for Computational Mathematics at the Flatiron Institute. Previously, he completed his PhD at the NYU Courant Institute of Mathematical Sciences, CILVR group, advised by Rajesh Ranganath and Thomas Wies. He works on deep generative models and machine learning in the sciences.

---

Tuesday, March 17th, 1:20pm ET
Zeming Lin (Biohub // Evolutionary Scale)

Title: Protein Language Models

Abstract: In this talk I’ll introduce protein language modeling, focusing on the ESM family of models that I’ve helped develop over my career. We’ll see how self-supervised training on millions of natural sequences uncovers the “grammar” of evolution, enabling models to infer structure, understand function, and guide design.

Bio: Zeming is a cofounder at EvolutionaryScale, which recently joined forces with Biohub, an institute focused on combining frontier biology research with frontier AI. Previously, he's done a PhD at NYU and worked at Meta as a research engineer, helping develop pytorch.

---

Tuesday, March 24th, 1:20pm ET
Elana Simon (Stanford University)

Title: InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoders

Abstract: Protein language models (PLMs) have demonstrated remarkable success in protein modeling and design, yet their internal mechanisms for predicting structure and function remain poorly understood. We present a systematic approach to extract and analyze interpretable features from PLMs using sparse autoencoders (SAEs). By training SAEs on embeddings from the PLM ESM-2, we identify thousands of human-interpretable features that correlate with biological concepts like binding sites, structural motifs, and functional domains. In contrast, examining individual neurons in ESM-2 reveals significantly less conceptual alignment, suggesting that PLMs represent most concepts in superposition. Beyond capturing known annotations, we show that ESM-2 learns coherent concepts that do not map onto existing annotations and propose a pipeline using language models to automatically interpret novel latent features learned by the SAEs. As practical applications, we demonstrate how these latent features can fill in missing annotations in protein databases and enable targeted steering of protein sequence generation. Our results demonstrate that PLMs encode rich, interpretable representations of protein biology and we propose a systematic framework to extract and analyze these latent features. In the process, we recover both known biology and potentially new protein motifs.

Bio: Elana Simon is a PhD student at Stanford advised by James Zou, working on understanding what machine learning models learn from biological sequences and structures. Previously, she was an ML engineer at Reverie Labs designing small-molecule cancer drugs and studied computer science at Harvard, where she worked with Debora Marks on protein language models. She also writes in-depth ML-biology analyses on her blog matmols and has been actively involved in research and advocacy for Fibrolamellar Hepatocellular Carcinoma.

---

Tuesday, April 28th, 1:20pm ET (tentative)
Sam Rodriques (FutureHouse, Edison Scientific)

Title: Building AI scientists

Bio: Sam Rodriques is an inventor and entrepreneur and the founder of FutureHouse, a research lab focused on building AI scientists, and Edison Scientific, which commercializes AI agents for scientific discovery. He was previously head of the Applied Biotechnology Lab at the Francis Crick Institute and earned his PhD at MIT. Named one of Time Magazine’s 100 most influential people in AI in 2025, his work spans accelerating biomedical discovery, engineering human biology, and developing new institutional models for scientific research.