Sbir phase II: a physics-based machine learning platform for crystal structure prediction of small drug molecules

Lavo Life Sciences Inc. received a $1.21 million Cooperative Agreement from the National Science Foundation (NSF) Technology, Innovation, and Partnerships (TIP) program (CFDA 47.084), awarded July 15, 2025, for completion by June 30, 2027. The awardee is developing a physics-based machine learning software platform designed to predict crystal structure formations of small drug molecules. The platform leverages artificial intelligence and advanced simulation techniques to identify viable solid-form properties of active pharmaceutical ingredients without requiring exhaustive laboratory screening, thereby reducing time and cost in pharmaceutical development and optimization.

This SBIR Phase II project directly addresses pharmaceutical industry challenges by enabling scientists to predict which crystal forms will emerge during drug manufacturing and to assess their efficacy and stability characteristics with greater confidence and speed. By automating the crystal structure prediction process, the software reduces development timelines for bringing pharmaceutical products to market, minimizes laboratory experimentation costs, and mitigates risks associated with unpredictable drug behavior or shortened shelf life. The innovation supports broader public health objectives by accelerating the commercialization of life-saving medicines while advancing computational chemistry capabilities in the pharmaceutical sector.

Abstract

The broader impact of this small business innovation research (sbir) phase II project is to improve the health and welfare of the public, advance scientific and technological understanding of pharmaceuticals, and to accelerate the development and commercialization of new pharmaceutical products. In the development of new medicines, considerable time and money is spent optimizing drug products to be taken orally in a pill or capsule. Failure to adequately optimize this solid form may result in a medicine that has a short shelf life or unpredictable behavior when administered to patients. Recent developments in computer simulations allow scientists to optimize the drug product while spending dramatically less time and money in the laboratory. The proposed project enables scientists to optimize and predict the characteristics of drug products in less time and with greater confidence. This directly reduces the time required to take life-saving medicines to market, helps pharmaceutical companies invest in curing a larger number of diseases, and takes to market the latest innovations in computer simulation. The proposed project is a software able to predict many of the solid-form properties of new and in-development drug products. The active pharmaceutical ingredient of a new drug may crystallize, or become solid, in many ways that are unknown beforehand. Certain crystal forms may be entirely ineffective, while others cure a disease as intended. To ensure a useful form is crystallized, pharmaceutical chemists search for all possible forms with thousands of lab experiments. Even so, they do not know if they have explored the breadth of all possible forms. This exhaustive screening process may take months and still result in an inferior product. The proposed project uses modern artificial intelligence and simulation techniques to predict which forms exist and give chemists useful guidance in how to obtain them in just days. The resulting technology will be highly automated and general-use for any new drug product intended for oral administration. The proposed project is made possible by innovations made during the phase 1 project. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the foundation's intellectual merit and broader impacts review criteria.- subawards are not planned for this award.