OnCo
ideasIdea

AI-designed proteins that grip the floppy parts of cancer drivers

Many cancer proteins have shapeless, flexible regions that drugs cannot hold on to. New protein design software may be able to invent binders that clamp them.

Deep-learning protein design (RFdiffusion, AlphaFold-based hallucination) has produced high-affinity binders to structured targets and, increasingly, to peptides and disordered segments. Intrinsically disordered regions of MYC, fusion oncoproteins and transcription factors are the classic undruggable surfaces. Designed binders could serve as degradation handles, intrabodies, or CAR and bispecific targeting domains rather than as drugs themselves.

Hypothesis
Designed miniproteins achieve nanomolar binding to at least one disordered oncoprotein region and, when fused to a degradation domain, deplete the target in cells.
Rationale
Design methods have crossed the threshold for structured epitopes and now handle conformational ensembles; the modality is intracellular expression or conjugation, not oral dosing, which relaxes the chemistry constraints.
What would test it
Design and test 100 binders per target region against MYC and one fusion oncoprotein, with biophysical validation and a cell-based degradation reporter; publish successes and failures for model improvement.
Maturity
speculative
Who has to act
research
Cost to try
Medium ($1M to $50M)
Years to first evidence
7
Bottlenecks it attacks

Connected

5top

Pages like this

not linked directly; found by shared links

technologies

1
AI-driven drug & target discovery

companies

2
Generate:BiomedicinesIsomorphic Labs

bottlenecks

2
AI that is built but not validated or deployedThe undruggable drivers