Bioptimus, a French artificial intelligence model startup, said today it has raised $41 million that will go toward building what it says will be the world’s first universal AI foundation model for biology research.

The round was led by Cathay Innovation with participation from existing investor Sofinnova Partners, a healthcare-focused venture capital firm that led the company’s seed funding round less than 12 months ago. Bpifrance joined the financing through its Large Venture Fund alongside Andera Partners, Hitachi Ventures, Boom Capital Ventures, Pomifer Capital and Sunrise. Today’s funding brings the total raised by the company to $76 million.

Bioptimus said that it intends to revolutionize the way that biology research will be done by using artificial intelligence foundation models that have access to all biological data and knowledge. Most biology research tends to remain within a single scope such as DNA, proteins, cells or tissues, which limits its potential impact.

To address the problem, the company released its first model, H-Optimus-0, the world’s largest foundation model for pathology. Launched in November, the model has 1.1 billion parameters and is trained on a proprietary dataset of several hundreds of millions of images extracted from more than 500,000 pathology slides across 4,000 clinical practices.

By bringing together as many different scales and modalities into the foundation model as possible, Bioptimus said, it can produce a biology model that can help accelerate research by providing a better model of reality. It does this by integrating data from every level including molecules, cells, tissues and organisms that are built from datasets from imaging, genetics and more.

According to Bioptimus, many AI models on the market also suffer from similarly limited scope. By making H-Optimus-0 as universal as possible, it can handle as many biology cases as researchers need and provide them answers at any resolution within their expertise.

“What we are building is not just a technological breakthrough; it’s a transformative tool for understanding biology in its full complexity,” said co-founder and Chief Executive Jean-Philippe Vert. “Essentially, it’s like the GPT of biology — but instead of generating text, we’re simulating biology.”

Vert said by modeling entire organisms down to the smallest molecules, the AI will provide the capability to empower researchers to predict disease outcomes in cellular systems and the response to treatments, allowing them to design therapies. Beyond pharmaceuticals, it could drive breakthroughs in understanding how gene expression affects cell division in cancer patients or enable biotechnology discoveries.

The company said the funding will be used to enhance its current AI platform to integrate even more diverse data sources and therapeutic areas. It’s also intended to enable the company to expand its critical datasets to refine and validate its models.

According to the company, H-Optimus-0 has outperformed all other pathology models in independent benchmarks, including the Harvard Medical School’s HEST program and the University of Leeds assessments. In these studies, the model showed it was particularly successful for predicting gene expression based on cell identification and capable of accurately subtyping ovarian cancer.

Photo: Pixabay