---
title: "MIT's Boltz-2 Ranks 2x More Drug Hits With Just 300 Labels"
subtitle: "A small dose of assay-specific labels teaches Boltz-2's affinity head to rank real hits far better, without retraining the giant structure model."
url: "https://alphasignal.ai/news/mit-s-boltz-2-ranks-2x-more-drug-hits-with-just-300-labels"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-09-21T09:02:04.000Z"
updated: "2026-09-25T01:28:27.808Z"
type: "Paper"
topics:
  - "Post Training"
  - "Benchmarks"
subtopics:
  - "Fine Tuning"
  - "Distillation"
upvotes: 99
source_url: "https://arxiv.org/abs/2609.24302"
comments: 0
truncated: true
---

# MIT's Boltz-2 Ranks 2x More Drug Hits With Just 300 Labels

> A small dose of assay-specific labels teaches Boltz-2's affinity head to rank real hits far better, without retraining the giant structure model.

# 300 assay labels sharpen Boltz-2 virtual screening

A new [preprint](https://arxiv.org/abs/2609.24302) reports that target-specific fine-tuning can improve Boltz-2 rankings with 300 binary assay labels. Across eight MF-PCBA targets, updating only the model’s affinity heads increased the number of actives in the top 1% by a geometric mean of 1.77 times and improved average precision by 2.14 times relative to the unmodified model.

Virtual-screening campaigns often rank millions of compounds while experimental budgets cover only a small fraction of the library. Assay-specific data can therefore matter more than broad benchmark performance: modest ranking changes determine how many confirmed actives reach the test plate.

## A broad model meets a narrow assay

Boltz-2, developed by MIT’s Jameel Clinic and Recursion, jointly predicts protein-ligand complex structure and binding affinity. Its affinity module was trained on about 5 million measurements, according to the [model overview](https://boltz.bio/boltz2).

On the retrospective MF-PCBA benchmark, the [Boltz-2 paper](https://jeremywohlwend.com/assets/boltz2.pdf) reports roughly twice the average precision of earlier machine-learning and docking baselines, plus an enrichment factor of 18.4 in the top 0.5%. An enrichment factor of 18.4 means that this narrow slice contains 18.4 times as many actives as random selection would yield.

Benchmark scores average performance across targets, while production screens depend on one protein construct, one assay protocol, and one chemical library. The new study tests whether a small set of measurements from that exact assay can improve Boltz-2 without retraining its costly structural core.

## The trunk stays frozen

The adaptation keeps Boltz-2’s structure trunk fixed and updates only its affinity heads. The trunk generates learned representations of the protein-ligand complex; the heads convert those representations into binding scores. Freezing the trunk avoids updating most model parameters, although uncached compounds still require a forward pass to generate their representations.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [Adapting Boltz-2 with limited experimental activity data improves early enrichment in virtual screening](https://arxiv.org/abs/2609.24302)
- [Boltz-2: Joint Structure and Binding Affinity Prediction](https://boltz.bio/boltz2)
- [Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction](https://jeremywohlwend.com/assets/boltz2.pdf)
- [Boltzina: Efficient and Accurate Virtual Screening via Docking-Guided Binding Prediction with Boltz-2](https://arxiv.org/abs/2508.17555)
- [MIT and Recursion Release Boltz-2](https://ir.recursion.com/news-releases/news-release-details/mit-and-recursion-release-boltz-2-next-generation-ai-model)