Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Evo 2 AlphaFold3 Cancer Classifier
Chaining Evo 2 and AlphaFold3 to classify cancer variants, this talk details a pipeline producing tiered therapeutic guidance validated with 3D protein-drug interactions.
I built a variant-to-therapy guidance system by chaining Evo 2 (40B genomic foundation model) with AlphaFold3 for structural validation. The pipeline takes a cancer mutation as input and outputs a tiered therapeutic recommendation with mechanistic rationale, multi-modal biological scoring, and 3D protein-drug interaction validation.
The technical demo will walk through:
- How I fine-tuned Evo 2 embeddings for multi-class variant classification (essentiality, functionality, chromatin, regulatory impact)
- My custom scoring layer that converts sequence likelihoods into actionable therapeutic guidance
- The AlphaFold3 integration that validates every guide/drug-target pair at atomic resolution
- Live code walkthrough of the inference pipeline: VCF → Evo 2 → scoring → AlphaFold3 → JSON output
I’ll show the actual Python code, the model architecture, and walk through a real BRCA1 ovarian cancer case where the system correctly identified synthetic lethality despite low essentiality scores.