iiinova — where your digital self keeps evolving

Minji Lee

@m1nj2.bsky.social

PhD student at Columbia CS. prev. Bachelor at KAIST CS haewonc.github.io

Minji Lee @m1nj2.bsky.social · 21/04/2026

Huge thanks to Colin, Minkyu, and Aymen for all the help, and @alisiafadini.bsky.social, @moalquraishi.bsky.social for the guidance!!

020

Minji Lee @m1nj2.bsky.social · 21/04/2026

arXiv: arxiv.org/abs/2604.18559 Code: github.com/aqlaboratory... All training, inference, and benchmarking data/code is open-sourced! 🧵 8/8

arxiv.org

ConforNets: Latents-Based Conformational Control in OpenFold3

Models from the AlphaFold (AF) family reliably predict one dominant conformation for most well-ordered proteins but struggle to capture biologically relevant alternate states. Several efforts have foc...

150

Minji Lee @m1nj2.bsky.social · 21/04/2026

These results suggest that AF3 models may learn (uncalibrated) implicit conformational landscapes, and that the right interface can unlock new capabilities without retraining. We're excited to explore interpretability, transferability, and applications of ConforNets. 🧵 7/8

110

Minji Lee @m1nj2.bsky.social · 21/04/2026

We train “conformational prompts” on 3 therapeutically important protein families and show strong induction of desired states: - GPCR active: 24% → 79% - Kinase DFG-out: 6% → 23% - Transporter outward-facing: 16% → 57% Built-in templates cause no observable shift. 🧵6/8

100

Minji Lee @m1nj2.bsky.social · 21/04/2026

Yet, the above benchmarks only ask whether an alternate state is observed among hundreds of samples. Can we prompt the model to consistently predict states that are rarely sampled? A "conformation prompt”? 🧵 5/8

110

Minji Lee @m1nj2.bsky.social · 21/04/2026

On multi-state benchmarks such as membrane transporters, ConforNets double recovery of alternate states from 24.3% → 51.1% over the OF3p baseline, at the cost of 3 GPU-minutes for a 400-residue protein. 🧵4/8

110

Minji Lee @m1nj2.bsky.social · 21/04/2026

2. Learn a desired state from one protein that can then be induced in many other proteins 🧵3/8

110

Minji Lee @m1nj2.bsky.social · 21/04/2026

Our key idea is to operate globally across Pairformer channels, instead of residues, to broadly sculpt conformational preferences. This makes it possible to: 1. Maximize unsupervised objectives like conformational diversity by coordinately training multiple ConforNets 🧵2/8

110

Minji Lee @m1nj2.bsky.social · 21/04/2026

We introduce ConforNets, a mechanism for conformational control in AlphaFold3 models - SoTA at producing diverse conformations on every multistate benchmark (N=104) - Novel capability: transfer state from one protein to another Outperforms BioEmu, ConforMix and AFsample3 🧵1/8

1419

Minji Lee @m1nj2.bsky.social · 21/04/2026

000