L1 Protein dynamics
Beyond static structure prediction, into the dynamics that actually govern function — and the physical invariants we can compress.
Life sciences
Protein dynamics, biological discovery, and reasoning over decades of published evidence.
Prediction without validation is publishing, not science. Our partners run the assays. Real measurements feed the next iteration.
protein dynamics literature reasoning closed-loop discovery
How we work
Models meet wet labs
Our life-sciences work is built around the conviction that prediction without validation is publishing, not science. We pair models with experimental partners who validate, push back, and feed real measurements back into the next iteration — and we ground every claim in literature evidence the team can trust.
prediction-only → publishing, not science
prediction + validation → science the next iteration can build on
Three lines of work
Where the science happens
L2 Literature reasoning
Reasoning systems that help biologists connect decades of published evidence to the experiment in front of them, with provenance attached.
L3 Closed-loop discovery
Predictions are paired with experimental partners who validate, push back, and feed real measurements back into the next iteration.
Evidence provenance
Decades of evidence, every claim traceable.
Each row is a real evidence type the literature surface indexes. The weight column is the contribution to the running claim, not a paper rating.
id year title kind weight
P-2024-AKT 1994 kinase substrate binding under conformational drift primary 0.84
R-2008-FLEX 2008 review · protein-flexibility datasets, 1994–2008 review 0.71
D-2017-MD-LG 2017 long-timescale molecular-dynamics ensemble release dataset 0.78
P-2021-CRYO 2021 cryo-EM evidence for an alternate hinge state primary 0.92
C-2025-CTRL 2025 control trial · in-vitro flexibility under mutation partner 0.88
Partner protocol
A model handoff a wet-lab partner will run.
Pre-registered. Confidence intervals attached. Success and failure criteria agreed before the experiment runs.
01
model
predict
predicted ensemble + confidence interval per residue
02
shared
spec
pre-registered assay · success and failure criteria
03
partner
measure
partner runs experiment, returns raw + processed
04
shared
fold-back
measurement enters the next iteration as evidence
From literature to lab
We work with biologists to make sense of decades of published evidence and link it to the experiment in front of them, with provenance the team can trust.
Models meet wet labs
Predictions are paired with experimental partners who validate, push back, and feed real measurements back into the next iteration.