RESEARCH

In MMLI 2.0, the overall goal of Thrust 1 is to develop foundational AI agents for discovery and synthesis of functional molecules, with a focus on three key areas:

1. A function- and synthesis-aware modular chemical language model (mCLM). mCLM tokenizes small molecules into function-infused and synthesis-friendly building blocks, akin to the way peptide/protein LLMs tokenize at the amino acid level. By design, the approach only generates chemical building blocks that can be iteratively assembled with MMLI’s robotic small molecule synthesis platform Molecule Maker Lab (MML). By doing so, mCLM will enable the prediction of new molecules of emergent functions derived from their modules, and all of the AI-generated molecules will be accessible via automated assembly by non-specialists.
2. A knowledge-augmented theme-specific LLM (themeLLM) for hypothesis generation and experimental design.
3. Innovative AI Agents with critical thinking. Notably, the development of other frontier AI tools such as generative AI will be performed within Thrusts 2-4, which further strengthens cross-thrust collaborations.

The AlphaSynthesis can be used for both manufacturing of target molecules with known functions (reverse synthesis) and discovery of molecules with new functions (forward synthesis). In both cases, identifying catalysts is critical to the success of implementing the synthetic routes. Therefore, the main goal of this Thrust is to develop new AI algorithms or approaches to discover and optimize catalysts that are required to implement the synthetic routes designed by AlphaSynthesis. In addition, these tools will be applied toward the realization of the four experimental endeavors and the catalysts required for the manufacturing of target molecules and materials in other Thrusts.

Inspired by exciting recent advances in molecule synthesis, the main goals of this Thrust are (1) to validate and test the AI-driven synthesis planning tool (Thrust 1) and the related AI tools for catalyst design and optimization (Thrust 2), and (2) to develop AI-driven tools for end-to-end optimization of entire synthesis plans. We will explore the retrosynthesis tool of AlphaSynthesis (reverse synthesis) to design synthetic routes for manufacturing of the following three target molecules with known functions: C2’epi amphotericin B (a novel potent and non-toxic antifungal drug candidate), artemisinin (a critical antimalaria drug), and Millad NX 8000 (an environmentally advantageous colorless, odorless thermoplastic clarifier for polypropylene). These routes will explore both chemical catalysis and biological catalysis.

Organic photovoltaics (OPV) are next-generation devices for harvesting renewable energy that are becomingly increasingly desired over traditional silicon solar cells. The two primary challenges preventing the widespread adoption of OPV relative to silicon-based solar cells are inferior power conversion efficiency and apparent instability to sunlight. Therefore, the main goal of this Thrust is to develop new machine learning tools to guide the discovery of highly efficient and indefinitely stable organic photovoltaics via automated synthesis, manufacture, and testing at the device level.