#USA #Cambridge #Drug Discovery #Insilico Medicine #Liquid AI

A Game-Changing Collaboration in Drug Discovery

A New Era for Pharmaceutical Research

In an exciting development for the pharmaceutical industry, Insilico Medicine has joined forces with Liquid AI to create lightweight scientific foundation models specifically designed for drug discovery. This groundbreaking partnership leverages advanced AI technology to address the pressing challenges that pharmaceutical companies face in efficiently utilizing artificial intelligence while safeguarding proprietary data.

Addressing Industry Challenges

One of the significant hurdles in drug development today is the need for pharmaceutical companies to harness powerful AI capabilities without exposing sensitive data related to molecules, assays, and targets to external cloud services. The collaboration between Insilico Medicine and Liquid AI aims to tackle this issue directly. By integrating Liquid AI's efficient Liquid Foundation Model (LFM) architecture with Insilico's MMAI Gym—a training platform that includes over 1,000 benchmarks tailored for pharmaceutical use—the partnership demonstrates that effective solutions can be deployed on-premises, yielding competitive results across various drug discovery tasks from a singular system.

The Breakthrough Model: LFM2-2.6B-MMAI

The fruit of this collaboration is the launch of LFM2-2.6B-MMAI (v0.2.1), a model that stands out for its compact size yet high performance. At just 2.6 billion parameters, this model has achieved state-of-the-art success across multiple drug discovery benchmarks, outperforming larger models in several categories. This advancement signifies a shift towards more efficient model designs, grounded in the principle that well-architected models can achieve excellent results without the need for excessive scale.

Diverse Capabilities of the LFM2-2.6B-MMAI

The model effectively covers a broad spectrum of the drug discovery process, including:

• - Property Prediction: Outperformed TxGemma-27B, a model over ten times its size, achieving top results in pharmacokinetics and toxicology across 13 out of 22 tasks.
• - Molecular Optimization: Demonstrated an astounding success rate of up to 98.8% on the industry-standard MuMO-Instruct benchmarks.
• - Affinity Prediction: Achieved superior correlation scores compared to leading frontier models like GPT-5.1 and Claude Opus 4.5, based on a robust internal benchmark featuring 2.5 million experimental measurements across 689 protein targets.
• - Chemical Reasoning: Showed commendable abilities in functional group reasoning and retrosynthesis planning, critical areas in medicinal chemistry.

These capabilities equip pharmaceutical companies to enhance their workflows, particularly in areas demanding high-frequency ADMET screening and lead optimization in medicinal chemistry.

Vision for the Future

Alex Zhavoronkov, the CEO of Insilico Medicine, expressed enthusiasm about the partnership, highlighting how lightweight liquid foundation models could transform the efficiency of scientific research. They pave the way for more scientists and researchers in the pharmaceutical landscape to expedite discovery timelines and provide better health outcomes for patients.

About Liquid AI and Insilico Medicine

Liquid AI focuses on building efficient Liquid Foundation Models based on dynamical systems and signal processing, ensuring their models can be deployed even in environments with limited resources. For more information, visit liquid.ai.

Insilico Medicine, on the other hand, stands at the forefront of biotechnology by utilizing AI for drug development across various conditions including cancer and aging-related diseases. Their insights and innovative technologies are paving the way for future breakthroughs in the industry. For further details, visit insilico.com.

Conclusion

The alliance between Insilico Medicine and Liquid AI marks a significant advancement in the pharmaceutical field, demonstrating that highly efficient and powerful AI models can be achieved at reduced costs without compromising quality. As this partnership unfolds, it promises to reshape the landscape of drug discovery, ultimately benefiting scientists and patients alike.