SBIR-STTR Award

The discovery of advanced materials and molecules is difficult and largely an experimental endeavor. Computers are used to simulate materials and molecules but rarely to design them. We aim to demonstrate that novel materials and molecules can be computationally “evolved” using Genetic Programming, and we will develop: (1) Atomically precise membranes that would separate molecules at a fraction of the energy used currently in the chemical separations industry ($11 billion/year market); (2) Green catalysts that transform inexpensive starting materials into pure, valuable products; (3) New therapeutics that wrap around and disable any target protein; and (4) Robust, environmental sensors for any target small molecule. We have developed a new software development environment called “Cando” that enables the development of software for designing materials and molecules, primarily based on a new macromolecular chemistry called “Molecular Lego” but is generally applicable to the design of all organic molecules. In phase I of the project, we will develop software for designing robust, atomically precise, single molecule thick membranes that display the ultimate in high selectivity and high flux for the separation of hydrocarbons, gasses, liquids and small molecules. The software will be used to design membranes for specific separations and then we will contract their synthesis. We will develop markets for these membranes in the $11 billion market for chemical separations. In the phase I, we will also develop software for designing small molecule therapeutics to run “free-energy perturbation” (FEP) calculations. FEP calculations are heavily used in the pharmaceutical industry to assist in drug discovery. FEP calculations are extremely difficult for drug discovery chemists to carry out and we will develop a revolutionary new user interface and backend to achieve this. There is currently only one commercial provider for FEP simulation software and this software is a ~ $50 million/year market. FEP simulation software represents a “target of opportunity” for the small business that we are starting (possibly named “Candesign”) and we already have interest from one major pharmaceutical company. In Phase II, and beyond, we will develop software within Cando for designing macromolecular catalysts and new therapeutics based on Molecular Lego technology. We will then transition into a Molecular Lego synthesis and design company that markets advanced functional materials and macromolecules that rival biological materials in efficiency and capability but are robust and engineer-able. The company that we are starting (potentially named “Candesign”) will become the leader in “Organic Bottom-up Nanotechnology”.

The state of computational chemistry software is in fragmented. Excellent and trusted academic software, which contains advanced molecular simulation code, exists but is seldom used in industry. Industrial software is closed, expensive and has not proven to be as good a foundation as academic software for developing advanced methods where machine learning can be used to design new molecules and materials. A software environment that combined the polish and convenience of commercial software with the open-source and community of academic software would serve as the foundation for the development of new computational methods. The Schafmeister group, has developed and with ThirdLaw LLC, will commercialize a revolutionary and open-source, computational chemistry programming environment called “Cando”. Within Cando, software applications (Cando-apps) are being developed that apply machine learning to computationally “evolve” functional macromolecules and molecular materials. Cando is an open source, LLVM based programming environment incorporating a compiler for a powerful dynamic programming language tailored for solving problems in molecular design. Cando builds on Unix, MacOS and Amazon Web Services machines and uses a client/server, Jupyter-notebooks based user interface, including molecular visualization and molecular sketching for chemists to interact with. Cando is designed to enable the computational, “evolutionary” design of Molecular Lego, a revolutionary new chemical building block and synthetic assembly method that enable the construction of nanometer scale structures with precise control over their three-dimensional structures. By applying computational evolutionary design to designing Molecular Lego nanostructures, we will create new molecules that act as new medicines, create channels that can filter gasses and small molecules to purify them from mixtures, and create molecular devices that build other molecules like (catalysts). Cando is very general and can be applied to many problems in computational chemistry. In the Phase I project we developed the Cando-app “fep-run” application to carry out free-energy perturbation calculations for drug discovery. We also laid the foundation for the Cando-app “membrane-designer”: to use machine learning to design atomically precise membranes for filtration using Molecular Lego nanostructures. The “fep-run” application is of great interest to pharmaceutical companies and so we focused on getting the prototype completed because it is the shortest path to developing a commercial revenue stream. Cando is open-source using an LGPL 2.0 license and so it will be distributed widely to academic and industrial users. In the Phase II project we will collaborate with pharmaceutical companies to tailor the “fep-run” application to their needs and develop an Amazon Web Services cloud based solution that we will resell to drug discovery companies to efficiently carry out valuable free-energy perturbation calculations at the lowest cost possible. We will also improve Cando and develop web-based training materials. Further - we will develop Cando applications that will use machine learning algorithms to design atomically precise catalysts, membranes and therapeutics.