Tahoe Therapeutics Inc., a startup using artificial intelligence to speed up medical research, today announced that it has closed a $30 million investment.

Amplify Partners led the funding round. It was joined by Databricks’ venture capital arm and a half dozen other institutional blockers. Tahoe is now valued at $120 million.

Oncological research could be carried out more efficiently if scientists had a way to reliably simulate how individual cancer cells behave in different situations. Using such technology, clinicians could quickly test how a cancer cell reacts to different drugs and identify the medicine that is most effective. San Francisco-based Tahoe is seeking to develop AI models that can run such simulations.

The main perquisite to building an AI model that can perform a task well is high-quality training data. Tahoe has built a platform called Mosaic that can generate training dataset for cell simulation models. In February, the company open-sourced a dataset called Tahoe-100M that was created using the platform.

Mosaic generates data using cancer cell lines, or cancer cells stored in a lab. Tahoe organizes those cells in configurations dubbed cell villages that help reduce data quality issues. The company exposes the cells to small molecules, which are molecules that often form the basis of medicines, and collects data on how the cells react.

Researchers historically collected cell behavior measurements using a method called single-cell transcriptomics. The technology can provide high-level data about the behavior of large cell groups, but provides limited information on individual cells. According to Tahoe, Mosaic uses a newer method called scRNA-seq that provides significantly more granular data than single-cell transcriptomics.

Tahoe-100M, the dataset that the company released in February, includes more than 100 million data points about cell behavior. The company produced the information by using Mosaic to collected information on over 150 million cells. According to Tahoe, its scientists collected 1.4 trillion data points during the project and then filtered them to leave only the highest-quality measurements.

In July, the company released a tool called TahoeDive to help researchers navigate Tahoe-100M. It provides the ability to run queries and visualize the results with natural language prompts. Under the hood, TahoeDive is powered by an AI agent that uses multiple language models to process user instructions.

Tahoe will use its new funding to create an AI training dataset that will be ten times larger than Tahoe-100M. In parallel, the company is working on AI models that will use the training data it produces to help researchers simulate cancer cells.

“Building Tahoe-100M required us to invent new ways to generate single-cell data,” said co-founder and Chief Executive Officer Nima Alidoust. “Now, we’re applying that superpower to go 10x further. This next phase is about using these massive datasets to bring about the GPT moment for AI models of human cells, translating insights to clinical readouts, and developing new medicines with much lower clinical failure rates.”

Down the road, Tahoe plans to team up with a drugmaker or AI company to support its training data production efforts. Additionally, the company hopes to collaborate with its future partner on the development of new medicines.

Photo: Unsplash