October 14, 2026
As we prepare to bring Lilly TuneLab's AI models directly into the CDD platform giving eligible companies access to predictive models trained on decades of Lilly research, we want the CDD community to understand how the program works, why we chose to partner with Lilly, and what it takes to obtain real value from models like these.
TuneLab is a collaborative AI platform that supports drug discovery and development. It provides access to powerful AI/ML models trained on decades of Lilly research, expertise and proprietary data. TuneLab members contribute data via a third-party hosted and operated federated learning infrastructure designed to be privacy preserving, to enable AI/ML drug discovery and development collaboration.
What is federated learning?
Federated learning is a way to train a shared machine learning model across several organizations
without pooling their proprietary data in one central database. Instead of bringing the data to the model, the model travels to the data.
In a typical setup, each participant keeps its data on its own "node"; A shared model is sent to each node and trained locally. Only what the model learned, in the form of model updates, goes to a central aggregator; the underlying data stays with its owner. The aggregator combines updates from all
participants into an improved shared model and sends it back out to each node.
For companies participating through CDD Vault, your company’s compound structures, antibody
sequences, and assay results are not shared with Lilly or other participants.
For drug discovery, that design matters. Predictive models generally improve with more data, and
federated learning is designed to increase the effective volume of training data while protecting the
confidentiality of each participant's data (JCIM).
What is Lilly TuneLab?
Eli Lilly and Company (Lilly) launched TuneLab in September 2025 to give biotech companies access to AI/ML models trained on Lilly's own research data (Lilly). According to Lilly, the first release draws on drug disposition, safety and preclinical data from hundreds of thousands of unique molecules, data Lilly estimates cost more than $1 billion to obtain. Today that effort is represented by 130+ member companies who joined Lilly TuneLab. (as of October 2026)
In Lilly’s September 9, 2025 launch announcement, Lilly CSO Daniel Skovronsky, M.D., PhD., described the goal plainly: Lilly TuneLab "was created to be an equalizer so that smaller companies can access some of the same AI capabilities used every day by Lilly scientists."
The CDD Vault integration will initially bring access to two core models:
ChemLab: Predict small-molecule ADMET properties across 30+ common endpoints, including
solubility, permeability, metabolic stability and safety-related properties.
AbLab: Explore developability predictions for conventional monoclonal antibodies across 17+ endpoints, including properties relevant to stability, aggregation and solubility.
On October 6, Lilly TuneLab added new models to its platform including an in vivo pk multi-species
model, NanoLab for VHH developability predictions, an antibody-antigen binding affinity model to predict how a specific mutation changes binding affinity between an antibody and its antigen, and an
immunogenicity (MHC-II peptide presentation) model to predict which peptides from an antibody
sequence are likely to be presented on MHC class II.
Why we brought TuneLab into CDD Vault?
We partnered with Lilly to bring the models to the place you and many biotech teams already do their
science. A prediction means the most when it sits beside everything else you know about a molecule.
In CDD Vault, each compound or biologic is registered once and becomes the anchor for every result that follows: biochemical and in vitro assays, DMPK, in vivo studies and profiling panels, along with the ELN entries that record how each experiment was run. Because those relationships are captured as structured data from the start, your team can ask questions across the whole picture without stitching together spreadsheets and exports.
That foundation is what makes this partnership meaningful for our customers:
● Predictions in context: Opted-in teams will be able to use select Lilly models within their existing
scientific workflows, so a predicted property profile can be weighed against the experimental
results already on record for the same molecule.
● Data that's ready to contribute: TuneLab asks participants to contribute training data. Assay
results that are already structured and linked to registered molecules are a far stronger starting
point than data that first has to be located, cleaned and reformatted.
● Fewer handoffs: Using an external model usually means exporting structures, running
predictions somewhere else and bringing the results back. Bringing TuneLab into CDD Vault is
meant to take those extra steps off your plate.
One CDD customer described at our Boston User Group Meeting how it combined TuneLab predictions with docking, lipophilicity calculations and medicinal chemistry judgment to prioritize compounds for synthesis. Individual results will vary.
Collaboration has been at the heart of CDD since its founding in 2004. As CDD CEO Barry Bunin put it when the partnership was announced: "We believe that solving the most complex challenges in drug discovery will depend on innovative collaboration models that provide broad access to research data and empower chemists and biologists to make informed, data-driven decisions."
Coming soon
The integration will be live tomorrow Thursday Oct 15. Join our upcoming webinar to see how ChemLab and AbLab work in CDD Vault, learn what participation involves, and bring your questions to the Lilly and CDD teams.. If you want to be among the first to see it, register here.
Lilly TuneLab membership is subject to eligibility criteria and the TuneLab terms of use, including data
contribution requirements, and is available only in supported regions. Model availability may change.
TuneLab predictions are intended to inform scientific decisions, not replace experimental data, and are
not intended for regulatory submissions.