Six months after emerging from stealth with $50 million in investment, Latent Labs developed a web-based AI model for programming biology.
CEO and founder Simon Kohl, a scientist who co-led DeepMind’s AlphaFold protein design team, says Latent Labs’ model achieved state-of-the-art on different metrics when testing its proteins in a real lab.
In AI, state-of-the-art (SOTA) refers to the industry’s top performance on a task.
He told TechCrunch, “We have computational ways of assessing how good the designs are,” adding that most of the model’s proteins will work in the lab.
LatentX, the company’s core biology model, lets academic institutions, biotech startups, and pharmaceutical corporations create novel proteins in their browser using natural language.
LatentX designs nanobodies and antibodies with precise atomic architectures that go beyond nature. Rarely can this method accelerate therapeutic development.
Kohl says LatentX’s ability to develop new proteins differentiates it from AlphaFold.
The Alpha fold model predicts protein structure. He said it lets you view existing structures but not produce new proteins.
Latent Labs licenses its model to other organizations, unlike AI-driven drug discovery startups like Xaira, Recursion, and DeepMind spinout Isomorphic Labs, which generate exclusive medications.
“Not every company can build their own AI models, infrastructure, and teams,” Kohl added.
LatentX is free, but Kohl said the company will charge for additional features as they are offered.
Chai Discovery and EvolutionaryScale also offer open-sourced AI drug discovery models.
Radical Ventures, Sofinnova Partners, Google’s Chief Scientist Jeff Dean, Anthropic’s Dario Amodei, and Eleven Labs’ Mati Staniszewski fund Latent Labs.
