Lilly and BigHat Target AI's Blind Spot in Drug Discovery

SAN MATEO, CA – January 07, 2026 – In a significant move to enhance the power of artificial intelligence in medicine, BigHat Biosciences announced today an expanded collaboration with pharmaceutical giant Eli Lilly and Company. The partnership will leverage BigHat’s advanced AI and automated laboratory platform to address a critical weakness in modern drug discovery, aiming to create more robust and reliable AI models for developing next-generation biologic therapies.

This new project focuses on advancing Lilly’s TuneLab, an AI platform designed to give biotechnology companies access to sophisticated drug discovery models. The collaboration builds on an existing relationship and signals a deeper commitment from both companies to solve one of the most pressing challenges in computational biology: the 'generalizability' of AI models.

The Generalizability Gap: AI's Critical Bottleneck

Artificial intelligence has long promised to revolutionize the slow and costly process of developing new medicines. However, a persistent hurdle has limited its full potential, particularly in the complex world of antibody therapeutics. The AI models used to predict whether a potential antibody drug will be safe, stable, and effective often struggle when presented with new and diverse types of molecules.

This limitation, known as a lack of generalizability, is at the heart of the new Lilly-BigHat collaboration. Peyton Greenside, CEO and Co-Founder of BigHat, articulated the problem in the company's announcement. “Current AI/ML models for antibody developability have limited ability to generalize to new sequences, largely because they are trained on limited formats and datasets of inconsistent quality,” she recognized. “As a result, they are primarily used for prioritization amongst hits, rather than to meaningfully improve developability or inform early, high-impact development decisions.”

This gap means that while AI can help sift through known candidates, it often falls short of guiding the creation of truly novel therapies or confidently predicting the success of molecules that fall outside its training experience. This forces drug developers to be more conservative, potentially shelving innovative but riskier drug targets and modalities that could lead to breakthroughs for patients with serious unmet needs.

A Strategic Alliance for Smarter Data

To solve the generalizability problem, AI models need to be trained on vast, diverse, and exceptionally high-quality data. This is precisely where BigHat's technology comes into play. The company will deploy its Milliner™ platform to rapidly generate these crucial datasets for Lilly's TuneLab.

Milliner is not just a set of algorithms; it is a fully integrated system that marries state-of-the-art machine learning with a high-speed, automated wet lab. This platform operates in rapid “design-build-test” cycles. First, AI models design hundreds of novel antibody variants. Then, the automated lab synthesizes and experimentally characterizes these molecules, measuring key properties like stability, solubility, and target binding. The real-world experimental data is then fed back into the AI models, creating a continuous learning loop that refines their predictive power with each cycle.

Overseeing this entire process is RADS (Reccy Antibody Design Studio), BigHat's custom-built operating system that orchestrates everything from robotic workflows to data processing and model training. This capability to generate and characterize hundreds of antibodies weekly allows BigHat to create the exact kind of diverse, high-quality data needed to build a more generalizable foundation model for antibody developability—a core objective of the TuneLab project.

Lilly's AI Playbook: A Bet on External Innovation

This expanded partnership is a key part of Eli Lilly’s broader strategy to position itself at the forefront of AI-driven pharmaceutical research. Lilly launched the TuneLab platform in September 2025, backing it with proprietary data from its own research efforts valued at over $1 billion. The platform's goal is to democratize access to powerful AI tools, allowing smaller biotech firms to leverage Lilly's extensive knowledge base.

TuneLab operates on a federated learning model, a sophisticated approach that allows partners to help train and improve the central AI models without ever directly sharing their own proprietary data. This creates a powerful ecosystem where the collective intelligence grows, benefiting all participants while maintaining data security. Before this latest announcement, Lilly had already brought roughly a dozen startups into the TuneLab ecosystem, including Insitro, Circle Pharma, and Firefly Bio.

This initiative is a cornerstone of Lilly's Catalyze360 program, which also provides strategic funding, lab facilities, and development expertise to promising early-stage companies. The relationship with BigHat is a prime example of this strategy in action. It began with a strategic partnership announced in April 2025, which included an equity investment from Lilly and a collaboration to design antibodies for up to two therapeutic programs. The success of that initial work has now paved the way for this deeper, more foundational collaboration focused on improving the very engine of AI-powered discovery.

From Lab Models to Lifesaving Medicines

Ultimately, the success of this collaboration will be measured not in datasets or algorithms, but in the new medicines it helps bring to patients. By creating AI models that can more confidently predict the success of early-stage drug candidates, Lilly and BigHat aim to de-risk the development process. This increased confidence could empower researchers to pursue challenging but high-impact targets for diseases that currently have few effective treatments.

The potential is exemplified by BigHat's own pipeline, which includes a next-generation antibody-drug conjugate (ADC) for gastrointestinal cancers and an avidity-driven T-cell engager (TCE) for solid tumors, both of which are projected to enter clinical trials in 2026. These complex, next-generation therapies are precisely the kinds of molecules that stand to benefit most from more intelligent and predictive design tools.

By tackling the fundamental challenge of AI generalizability, the partnership between Lilly and BigHat represents more than just a technological advancement. It is a strategic effort to shorten development timelines, reduce the high attrition rates that plague drug discovery, and ultimately accelerate the journey of innovative therapies from a computer model to a patient's bedside.