Foundations

From "what is a protein" to reading a PDB structure critically — delivered as the series "Structural Biology from Zero."

1. Part 1Published

   Structural Biology from Zero · 1: Why You Can't See a Protein

   Proteins are far smaller than the diffraction limit of visible light. Structural biology therefore reconstructs their atomic structures from physical measurements.

2. Part 2Published

   Structural Biology from Zero · 2: The Twenty Amino Acids and the Backbone

   Every protein you are made of is spelled in a 20-letter alphabet strung along one repetitive, flat-linked chain. Learn the letters, the bond that joins them, and the two angles that decide almost everything about the overall shape.

3. Part 3Published

   Structural Biology from Zero · 3: How a Chain Finds Its Shape

   Protein folding reflects many weak interactions, solvent thermodynamics, and an energy landscape that favors productive routes toward the native ensemble.

4. Part 4Published

   Structural Biology from Zero · 4: The Levels of Structure

   A protein is organized on four scales at once — the sequence, the local hydrogen-bonded motifs, the folded domain, and the assembled machine. Each level follows from the one below by the same physics, and together they are the vocabulary the rest of structural biology is written in.

5. Part 5Published

   Structural Biology from Zero · 5: Enzymes — Structure as Catalysis

   Some enzymes accelerate reactions by factors approaching a hundred quadrillion through active-site geometry, electrostatics, and preferential stabilization of transition states.

6. Part 6Published

   Structural Biology from Zero · 6: Antibodies and the Immune System

   Antibody repertoires combine a conserved immunoglobulin fold with variable binding loops, gene recombination, and selection to recognize a broad range of molecular surfaces.

7. Part 7Published

   Structural Biology from Zero · 7: Seeing Molecules — X-ray, Cryo-EM, and NMR

   X-ray crystallography, cryo-electron microscopy, and NMR convert different physical measurements into molecular structures, each with characteristic strengths and limitations.

8. Part 8Planned

   Reading a Structure Critically

The bridge a physicist needs before proteins meet transformers. No prior deep learning assumed.

1. Article 1Planned

   Tensors, Gradients, and What "Learning" Means

2. Article 2Planned

   The Transformer from Scratch

3. Article 3Planned

   Embeddings and Representation Learning

4. Article 4Planned

   Variational Inference and the VAE

5. Article 5Planned

   Training Dynamics: Optimization, Regularization, Scaling

6. Article 6Planned

   Evaluation Done Right

The PLM landscape, model by model — the published overview is the pillar; each model gets a deep dive.

1. OverviewPublished

   Protein Language Models: Learning from the Sequence

   How transformers trained on millions of protein sequences learned to encode structure, function, and evolutionary history — and why that's useful for drug design.

2. Part 1Planned

   ESM-2 in Depth

3. Part 2Planned

   ProtBERT and the BERT Lineage

4. Part 3Planned

   ProGen and Autoregressive Generation

5. Part 4Planned

   ESMFold: From Embeddings to Structure

6. Part 5Planned

   AlphaFold2: MSA and the Structure Module

7. Part 6Planned

   ProteinMPNN and the Inverse-Folding Problem

8. Part 7Planned

   A Comparative Review

Why molecules are graphs and why symmetry has to be built into the network.

1. Article 1Planned

   Graphs Are the Native Language of Molecules

2. Article 2Planned

   Message Passing and GNNs

3. Article 3Planned

   Why Equivariance Matters

4. Article 4Planned

   EGNN and GVP

5. Article 5Planned

   SE(3)-Equivariant Networks

6. Article 6Planned

   Case Study: A Small Equivariant Model

The generative toolkit, from the SDE up to protein design.

1. Article 1Planned

   Diffusion Models from the SDE Up

2. Article 2Planned

   Flow Matching

3. Article 3Planned

   RL for Constrained Generation

4. Article 4Planned

   Bayesian Optimization and Active Learning

5. Article 5Planned

   Protein Design: RFdiffusion and the Design Problem

6. Article 6Planned

   Reproduction: One Generative Paper, End to End

Connecting the models to a molecule that survives contact with biology.

1. Article 1Planned

   Medicinal Chemistry for ML People

2. Article 2Planned

   ADMET and Why Molecules Fail

3. Article 3Planned

   PK/PD: What the Body Does to a Drug

4. Article 4Planned

   Toxicology and Safety

5. Article 5Planned

   From Target to Clinic

A weekly loop: read a stack of papers, find one real gap, sketch one experiment.

1. Article 1Planned

   Research Notes — a paper, a gap, an experiment (weekly)