Peer-reviewedCC BY 4.0Molecular Systems Biology
Validation and analysis of 12,000 AI-driven CAR-T designs in the Bits to Binders competition
35 affiliations
- 1Department of Molecular Biosciences, University of Texas at Austin, Austin, TX, USA
- 2The BioML Society, Austin, TX, USA
- 3LifEngine Animal Health (LEAH) Laboratories Incorporated, Minneapolis, MN, USA
- 4Department of Materials, Imperial College London, London, United Kingdom
- 5University of Texas at Austin, Austin, TX, USA
- 6Imperial College London, London, UK
- 7Monash University, Melbourne, VIC, Australia
- 8University of Toronto, Toronto, ON, Canada
- 9Institute for Drug Discovery, Leipzig University, Leipzig, Germany
- 10EPFL, Lausanne, Switzerland
- 11University of Oldenburg, Oldenburg, Germany
- 12Stockholm University, Stockholm, Sweden
- 13Czech Technical University in Prague, Prague, Czechia
- 14University of California San Diego, San Diego, CA, USA
- 15University of Oxford, Oxford, UK
- 16University of Texas Southwestern Medical Center, Dallas, TX, USA
- 17Virtue Therapeutics, Gurugram, India
- 18University of Illinois Urbana-Champaign, Champaign, IL, USA
- 19McGill University, Montreal, QC, Canada
- 20Sungkyunkwan University, Suwon, South Korea
- 21Vanderbilt University, Nashville, TN, USA
- 22Uppsala University, Uppsala, Sweden
- 23Gensaic, Cambridge, MA, USA
- 24Hebrew University of Jerusalem, Jerusalem, Israel
- 25University of Florida, Gainesville, FL, USA
- 26University of Washington, Seattle, WA, USA
- 27Purdue University, West Lafayette, IN, USA
- 28University of Tehran, Tehran, Iran
- 29Taras Shevchenko National University of Kyiv, Kyiv, Ukraine
- 30Myriad Genetics, Salt Lake City, UT, USA
- 31Harvard University, Cambridge, MA, USA
- 32Aibio AB, Klagshamn, Sweden
- 33Boltzmann Labs, Hyderabad, India
- 34National Institute of Chemistry, Ljubljana, Slovenia
- 35Seoul National University, Seoul, South Korea
Abstract
Artificial intelligence (AI) methods for proteins have advanced rapidly, improving structure prediction and design, particularly for de novo binders. However, most evaluations emphasize binding affinity rather than higher-order biological function. We present Bits to Binders , a global competition evaluating de novo binder design in the context of chimeric antigen receptor (CAR) T cells. Teams from 42 countries submitted 12,000 designs of 80-amino acid de novo minibinders targeting human CD20 as CAR binding domains. Designs were screened by pooled CAR-T proliferation, identifying 707 designs exhibiting significant CD20-specific enrichment, with team hit rates from 0.6% to 38.4%. Top-performing candidates were validated as individual constructs, measuring CD20-specific proliferation, expansion, cytokine production, and targeted cell lysis. We identified common design methodologies and factors correlated with DNA synthesis, expression, and target-specific T cell activation which nearly double the success rates when applied as a retrospective filter. We release this dataset as an open resource, with practical recommendations to support more effective AI-driven binder design.
Related
- Metabolic Competition-Driven Antitumor Immunity of glpX-Deficient Engineered Salmonella via Glucose Deprivation and STING Activation in Bladder Cancer
- A preliminar validation-driven experimental–numerical framework for vaginal delivery simulation
- New potassium binders: we need them—we do not need them
- Domain-resolved epitope mapping of B7-H3-targeting VHH-Fcs defines isoform-selective binders for antibody-drug conjugate development
- Deep Learning–driven Segmentation of Breast Cancer MR Imaging Using a Hybrid ResNet–Transformer Architecture and Cross-dataset Validation
BibTeX
@article{kosonocky2026validation,
title = {{Validation and analysis of 12,000 AI-driven CAR-T designs in the Bits to Binders competition}},
author = {Clayton W. Kosonocky and Alex M. Abel and Aaron L. Feller and Amanda Cifuentes Rieffer and Phillip R. Woolley and Jakub Lála and Delbert S. Barth and Tynan Gardner and Bits to Binders Competitors and Daniel Acosta and Stefano Angioletti‐Uberti and Vinayak Annapure and Neil Anthony and Rana A Barghout and Max Beining and Parth Bibekar and Dionessa Biton and Patrick Bryant and Anton Bushuiev and Cianna N. Calia and Ava Chan and Jie Chen and Akshay Chenna and Nicole Chiang and Vivian Chu and Joseph Clark and Alia Clark‐ElSayed and Elliot Cole and Qian Cong and Shreyasi Das and Tanner Dean and Tyler Derr and Alejandro Diaz and Jesse Durham and Moritz Ertelt and Antonio Fonseca and Ora Schueler‐Furman and Tynan Gardner and Jokent T. Gaza and Nathaniel Greenwood and Rui Guo and Abel Gurung and Vivian Haas and Seyed Alireza Hashemi and Dieter Hoffmann and Sofiia Hoian and Corey Howe and Calvin XiaoYang Hu and Jinling Huang and Nader Ibrahim and Satoshi Ishida and Ansar Ahmad Javed and Aswini Javvadi and Roman Jerala and Daksh Joshi and Even Kiely and Haelyn Kim and Diego Kleiman and Sarah Knapp and Huan Koh},
journal = {Molecular Systems Biology},
year = {2026},
month = oct,
doi = {10.1038/s44320-026-00246-1},
url = {https://link.springer.com/10.1038/s44320-026-00246-1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Validation and analysis of 12,000 AI-driven CAR-T designs in the Bits to Binders competition
AU - Clayton W. Kosonocky
AU - Alex M. Abel
AU - Aaron L. Feller
AU - Amanda Cifuentes Rieffer
AU - Phillip R. Woolley
AU - Jakub Lála
AU - Delbert S. Barth
AU - Tynan Gardner
AU - Bits to Binders Competitors
AU - Daniel Acosta
AU - Stefano Angioletti‐Uberti
AU - Vinayak Annapure
AU - Neil Anthony
AU - Rana A Barghout
AU - Max Beining
AU - Parth Bibekar
AU - Dionessa Biton
AU - Patrick Bryant
AU - Anton Bushuiev
AU - Cianna N. Calia
AU - Ava Chan
AU - Jie Chen
AU - Akshay Chenna
AU - Nicole Chiang
AU - Vivian Chu
AU - Joseph Clark
AU - Alia Clark‐ElSayed
AU - Elliot Cole
AU - Qian Cong
AU - Shreyasi Das
AU - Tanner Dean
AU - Tyler Derr
AU - Alejandro Diaz
AU - Jesse Durham
AU - Moritz Ertelt
AU - Antonio Fonseca
AU - Ora Schueler‐Furman
AU - Tynan Gardner
AU - Jokent T. Gaza
AU - Nathaniel Greenwood
AU - Rui Guo
AU - Abel Gurung
AU - Vivian Haas
AU - Seyed Alireza Hashemi
AU - Dieter Hoffmann
AU - Sofiia Hoian
AU - Corey Howe
AU - Calvin XiaoYang Hu
AU - Jinling Huang
AU - Nader Ibrahim
AU - Satoshi Ishida
AU - Ansar Ahmad Javed
AU - Aswini Javvadi
AU - Roman Jerala
AU - Daksh Joshi
AU - Even Kiely
AU - Haelyn Kim
AU - Diego Kleiman
AU - Sarah Knapp
AU - Huan Koh
PY - 2026
DA - 2026/10/08
JO - Molecular Systems Biology
DO - 10.1038/s44320-026-00246-1
UR - https://link.springer.com/10.1038/s44320-026-00246-1
AB - Artificial intelligence (AI) methods for proteins have advanced rapidly, improving structure prediction and design, particularly for de novo binders. However, most evaluations emphasize binding affinity rather than higher-order biological function. We present Bits to Binders , a global competition evaluating de novo binder design in the context of chimeric antigen receptor (CAR) T cells. Teams from 42 countries submitted 12,000 designs of 80-amino acid de novo minibinders targeting human CD20 as CAR binding domains. Designs were screened by pooled CAR-T proliferation, identifying 707 designs exhibiting significant CD20-specific enrichment, with team hit rates from 0.6% to 38.4%. Top-performing candidates were validated as individual constructs, measuring CD20-specific proliferation, expansion, cytokine production, and targeted cell lysis. We identified common design methodologies and factors correlated with DNA synthesis, expression, and target-specific T cell activation which nearly double the success rates when applied as a retrospective filter. We release this dataset as an open resource, with practical recommendations to support more effective AI-driven binder design.
SN - 1744-4292
ER -