Introducing Rhizome

Today, Rhizome Research, Inc. comes out of stealth. We are an applied AI lab based out in Seattle tackling small molecule discovery. Our immediate goal is to build autonomous in silico drug designers. We believe that the "primitives" of drug discovery have yet to be laid and while the space for AI methods is crowded the technology is still nascent. We are taking a bottom-up approach to this problem, reasoning from first principles about the models and physics pipelines that are needed to help make drug discovery a tractable problem.

These are our fundamental beliefs that we are sharing with the community and are willing to be judged on:

1. Large pre-trained Graph Neural Networks (GNNs) are the future and the only tractable way to approach small molecule modeling.
2. ABFE and RBFE calculations are the most critical part of any physics-based screening pipeline. They are the most important methods to accelerate by orders of magnitude without destroying accuracy.
3. LLMs will continue to improve but there are fundamental limitations. Instead of focusing and improving what they can't do, we double down on where they excel – orchestration, code execution and analysis.
4. Ligand-Based toxicity models are not the future. In 2026 we release what we believe is.
5. RL has been underutilized in drug discovery.

These beliefs will be proven or disproved the hard way: in our customer's wet lab and the clinic.

Please see here for the launch of our Foundational model for small molecule discovery.

Very soon we aim to launch MolSim (our accelerated Physics pipeline) and will open source a multi-agent system for running Docking and Molecular Dynamics (MD) simulations. In 2026 we will integrate our agents with our full pipelines and share our first end-to-end autonomous discovery loop on a real world program. Over time we will also announce our partnerships and share wet lab data in H1-2026 from the different targets such as EZH2 and SARM1 as well as many others we have in pipelines.

Long term, we believe language, physics, and chemistry will collapse into a single massive multimodal model.