BioSim M2M: Molecules to Medicine

Project Information

biosim m2m

Grant agreement ID: 190181387

DOI

10.3030/190181387

Project closed

EC signature date 28 August 2023

Start date 1 September 2023

End date 28 February 2026

Funded under

• The European Innovation Council (EIC)

Total cost

€ 3 570 750,00

EU contribution

€ 2 499 525,00

Coordinated by

BIOSIMULYTICS LIMITED
Ireland

Periodic Reporting for period 1 - biosim m2m (BioSim M2M: Molecules to Medicine)

Reporting period: 2023-09-01 to 2024-08-31

Summary of the context and overall objectives of the project

Biosimulytics is developing breakthrough technology combining AI and Quantum technologies to boost the success rates of pharmaceutical R&D in getting from Molecules to Medicine (M2M). The Biosimulytics innovation is being applied to crystal structure prediction (CSP) for determining the most stable crystal structure or polymorph of a drug compound. Polymorphism is a persistent headache for material scientists in the pharmaceutical sector. Molecules very often have several polymorphs, each with different properties that can seriously impact on the effectiveness or even the safety of the end product medicine. Polymorph problems cause a myriad of headaches for pharma/biotech companies from product recalls to regulatory delays to patent challenges, all of which represent very costly risks for the industry. Existing state-of-the-art techniques for polymorph analysis require long and painstaking experimentation by material scientists with uncertain results which can add up to 2 years or more to the time it takes to bring a new medicine to market. Biosimulytics aims to dramatically reduce the time to find the most stable crystal structures of a candidate drug molecule by using predictive technology to guide lab scientists on where to focus their experimentation work for a right-first-time result. This can reduce the number of polymorph screening experiments from 400 to 40, saving huge time as well wasted scarce resources used in experimentation. Most importantly, the goal of Biosimulytics is to make Crystal Structure Prediction (CSP) technology available, affordable and usable by every pharma/biotech company, including small and emerging biotechs which make up 80% of the R&D pipeline of new drug molecules, by innovating in the use of AI and Quantum technologies to reduce a typical CSP cycle from 3 months down to 3 weeks or less, and scalable as a cloud-based solution. As a result, Biosimulytics will contribute to the Pharmaceutical Strategy for Europe by helping to ensure that patients have access to high quality, effective, and safe medicines.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

The Biosimulytics EIC project was successfully kicked-off on 1st Sept 2023. A new Project Manager was recruited and the project management systems were installed at the outset to professionally track and manage the project’s progress and performance. A plan for HR management and recruitment was initiated and the key position of a full-time CTO was successfully filled. We also successfully recruited two additional Computational Scientists/Chemists, a Business Development specialist, and two new Board Directors (50% of the new hires are female). The development tasks relating to the company’s proprietary neural network model were completed. These covered the development of the neural network to include geometry dependent charges, the optimization of the neural network training algorithm, the establishment of an active learning approach for on-the-fly training of the neural network and the acquisition of an optimal training dataset for organic drug molecules. Most importantly, we advanced our prototype solution for Crystal Structure Prediction (CSP) from its initial Alpha state at the start of the project to a new Beta version. The Beta version of the company’s prototype CSP solution was successfully developed in line with latest market intelligence from customers, partners and investors with excellent preliminary results on relevant test molecules from the CCDC Blind Test which show the ability to successfully identify the experimental crystal structure of the molecule within the Top 10 predicted structures in a comprehensive polymorph landscape mapping achieved within the targeted timeframe of less than 3 weeks for a CSP cycle. In addition, strong progress was achieved on technology partnerships for scale-up of the Biosimulytics CSP solution, especially with Viridien which is a global leader in high-performance computing which is headquartered in France and provides Biosimulytics with vast expertise, experience and resources for accelerating and scaling the CSP solution for the mass market in Life Sciences, and with a cybersecurity start-up Binarii Labs which is helping to ensure the best-in-class data security features are built into the CSP solution using blockchain technologies. Finally, we also successfully completed initial scoping exercises on the deployment of CSP technology for use cases on solvate and amorphous molecules together with launching customers.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

One of the key results from the project to date is that Biosimulytics has successfully achieved the granting of the patent from the European Patent Office on its proprietary search algorithm for crystal structures (CONFIGURATIONAL ENERGY CALCULATION AND CRYSTAL STRUCTURE PREDICTION - EPO Grant Award No. 3948877). The international patent application is pending award in US, Canada, China and Japan. The project results to date have also triggered the consideration of a new patent application for the company’s intellectual property (IP) on its proprietary AI forcefield with an intent to file the new patent application in 2025 subject to a review of our IP protection strategy. The preliminary results achieved with the latest Beta version of the prototype for the Biosimulytics CSP solution which has been developed in the framework of the project are extremely encouraging. The results show the ability to successfully identify the experimental crystal structure of a relevant benchmark drug molecule from the CCDC Blind Test within the Top 10 predicted structures in a comprehensive polymorph landscape mapping achieved within the timeframe of less than 3 weeks for a CSP cycle, which is beyond the state-of-the-art for CSP technology. In order to bring these results fully to market, there are several critical tasks which still need to be completed. Further development work on free energy calculations and the development of additional Machine Learning components for CSP to handle much larger and more flexible molecules are needed. We also need to containerize the solution for optimal scalability, including optimizing the software and hardware with our technology partners, and complete the end-to-end testing of the solution. In parallel, we will need to test and validate the solution through a series of pilot projects with our launching pharma/biotech customers in Europe and North America.