Validation and analysis of 12,000 AI-driven CAR-T designs in the Bits to Binders competition
A competition tested AI-designed CD20 binders in CAR-T cells and released its dataset.
TL;DR
- The Bits to Binders study reports testing 12,000 AI-designed binders targeting CD20 in CAR-T cells.
- The paper reports 707 designs with significant CD20-specific enrichment in the pooled screen.
- The associated repository makes the competition data and analysis code available.
The preprint describes a functional screen of submitted designs, moving evaluation beyond predicted binding toward measured cell responses. [1] [2]
Its authors report 707 designs with significant CD20-specific enrichment; the repository provides the associated data and code for inspection. [2] [3] [4]
Why it matters
For biotech research, the study illustrates how protein-design claims can be tested against functional assays and shared datasets; it does not establish clinical efficacy.
Editor's note
The cited manuscript is identified as a preprint and has not been peer reviewed.