Challenge — BioML Society

(Update from 9/2/25)

The BioML Challenge: Bits to Binders, organized by the University of Texas at Austin BioML Society, is a 5-week competition helping to advance the field of AI protein design. In this challenge, we gathered scientists from around the world to design and submit protein binders that cause immune cells to target and eliminate CD20+ tumors. Rather than test for just binding affinity, we tested these binders directly in the therapeutic modality to see how well AI protein design tools work towards a real-world scenario that could one day accelerate cell therapy prescription to days or weeks, rather than years.

We had 28 teams with competitors from around 40 different countries design and submit a combined 12,000 peptide binders using their open-source methods of choice. All of the designs and methods, both in silico and in vitro, will be made public, and all parties have agreed that no IP will be claimed on them to help further open science.

Testing 12,000 putative antigen binder designs was only possible thanks to LEAH Labs’ high-throughput T cell engineering screen, DNA synthesis from Twist, reagents from Lonza, ScaleReady, and VWR, and additional funding from the Center for Systems & Synthetic Biology at UT Austin.

The binders functioned in the context of CAR-T cells, serving as the binding domain that helps the CAR-T recognize the cancer antigen CD20. To validate them, we did two rounds of testing:

Round 1 (High-Throughput Screen)

The 12,000 CAR-T cell constructs were tested in a high-throughput pooled screen to measure how well the binders drove CAR-T cell proliferation in the presence of the target antigen CD20 compared to a control. This approach hijacks a necessary but not sufficient behavior of functional T cells—proliferation—to identify which binders engaged the antigen, signaled through their CAR, and became overrepresented at the population level

Round 2 (Individual Functional Assays)

The ten best performing sequences from Round 1 were selected to move on to a series of individual assays measuring many aspects of T cell biology including cytotoxicity against CD20+ tumors, cytokine release, proliferation, and expansion. We created an aggregate score from these assays that allowed us to give an overall rank to the designs.

—

Each design in the top ten succeeded at some aspect of T cell biology, but a few were successful in all of them! The teams with the top three designs were the Perez Lab Gators, Amigo Acids, and the Schoeder Lab. Each of these teams will be given an award.

We’re also giving a second award to the team with the highest success rate of submitted sequences in the Round 1 screen. This team was Nucleate UK with a whopping 38% hit rate, with the runnerups being BindingIllini (15%) and Furman Lab (11%).

Huge congrats to the winners! Each of the four winners will be given a 3D print of their winning binder, some stylish LEAH Labs swag, and, most importantly, bragging rights!

The full data release, scores, and analysis will be coming soon when the publication hits preprint (hopefully in October). Stay tuned!

Nature feature on the Bits to Binders competition

Event Format

Teams were tasked to design the antigen binding domain of a Chimeric Antigen Receptor (CAR) that engages the cancer antigen target CD20. The goal is that the antigen binders activate a CAR-T cell killing and proliferation response, resulting in a quantifiable increase in the designed antigen binders that function while the designed antigen binders that don't will do nothing. Binder designs were constrained to 80 amino acids (a function of DNA synthesis limitations) with the rest of the CAR (both CD28z and 41-BBz) remaining constant between designs.

The challenge kicked off with a one-day remote event on Saturday August 24th at 11am CST. This kickoff consisted of an announcement of the binding target, overview of design constraints and supporting data, presentations by community partners, and a tutorial for a baseline design workflow. Breakout rooms were provided for teams to work together and receive technical support.

Following the kickoff, teams had 5 weeks to submit their designs. Teams were encouraged to use existing tools and develop creative methods at each stage of the design process: design, optimization, and downselection. Each team submitted their protein sequences for synthesis and testing alongside a GitHub repository with their code and a brief writeup of their methods.

Testing and Evaluation

Submitted sequences were synthesized as DNA and the corresponding proteins experimentally evaluated using LEAH Laboratories’ CAR-T platform. Finalists were selected based on how well their predicted antigen binders activated CAR-T cells in a pooled competition assay against CD20+ cancer cells, and the finalists were then screened individually in a variety of functional assays. More details to follow in the upcoming preprint.

The top teams are invited to present their methods and receive a trophy for their work. A manuscript is currently being compiled that outlines the design approaches, winning sequences, and lessons learned. This will be accompanied by a dataset and methods release to help further open science.

Awards and Scientific Dissemination

Participant Agreement

By participating in this competition, competitors agree to the following terms:

Intellectual Property: No intellectual property (IP) rights will be claimed on the designed sequences. All designed sequences must be released to the public domain. The organizers and sponsors will not be claiming intellectual property rights to the designed sequences nor will they be commercializing them.
Open Source and Public Availability Requirement:
- All datasets, models, methods, and code developed or used during the competition should be made open source whenever possible.
- In cases where open sourcing is not feasible (like AlphaFold3 or ESM-3), competitors must ensure that the models and methods are well-documented and made publicly available. This includes providing comprehensive documentation and access to the models to allow for public use and verification.
- Update (8/17/24): We discourage teams from using closed-source and license-restrictive models such as AlphaFold3 and ESM-3. However, we cannot police this given the number of submissions we are dealing with, so by submitting your designs you agree that the burden of responsibility for all software used rests upon you, the submitting party, and not the organizers or sponsors.

Further, the participant represents that they have the right to use any software, models, tools or data that they have used or will use to produce any entry in the competition. Participant also represents that Participant (i) has not filed a patent application or submitted an invention disclosure to Participant’s employer or institution for the purpose of a patent application being filed, and (ii) is not aware of any invention disclosure or patent application, that attempts to protect or cover the sequence, sequences, molecule or molecules, or any method of using them, that Participant has submitted or will submit as an entry in the competition. Participant further represents that Participant will not in the future file such an application or submit such an invention disclosure to an employer or institution.

Ethical Disclaimer

This competition is intended solely for educational and research purposes. All participants are advised to ensure that their designs comply with relevant ethical guidelines and regulatory standards, and the organizers disclaim any responsibility for misuse of the resulting protein binders or described methods.