02 / TARGETS · MOLECULES · EVIDENCE
Which step does a model address, and what evidence is still needed?
AI-assisted drug discovery spans targets, molecular interactions, cellular responses, and candidate validation. Start with the available design, docking, and perturbation analysis coverage, then explore the supporting methods.
Start exploring01 / GET YOUR BEARINGS
Follow the questions in order, or go straight to the part that matters to your work.
Expand each stage to explore ↗Describe the research question, target context, and data sources before choosing what to predict or design.
When reading docking and design results, track structure provenance, interaction restraints, and the conditions under which they apply.
Relate molecular predictions to cellular perturbation data while distinguishing dose, time, cell type, and other experimental conditions.
Binding, activity, selectivity, and further validation address different questions. Interpret each conclusion within its experimental and data context.
02 / READING
A guide to BindCraft2 v1.0.4 changes in antibody interfaces, induced-fit and fold-switching filters, and large-system execution, with the limits of the reported computational examples.
Read the article → 4 min · Chinese bodyClaude coordinated a protein-design campaign that produced 318,484 computational designs and 211 tested candidates. Seventeen met the authors’ binding criteria; toxin neutralization, broad coverage and therapeutic efficacy remain unproven.
HADDOCK3 2026.10.1 fixes a silent failure in contactairs. This guide explains the selection-expression and byres changes, added tests, and which refinement or scoring jobs merit review.
03 / RESOURCES