Intellicule has received a $217,941 NIH SBIR Phase I grant to develop advanced deep-learning software for biomolecular modelling. This technology aims to accelerate novel drug discovery and advance precision medicine by improving cryo-EM analysis
Intellicule, a Purdue University-affiliated software startup, has received a significant financial boost with a $217,941 Small Business Innovation Research (SBIR) Phase I grant from the National Institutes of Health (NIH).
The funding will be used to develop state-of-the-art software intended to accelerate the discovery of novel drugs and advance the field of precision medicine. The technology focuses on enhancing the analysis of biomolecular structures captured using cryogenic-electron microscopy (cryo-EM), a widely utilised technique in structural biology.
Accelerating novel drug development: Biomolecular modelling
The primary goal of the grant-supported project is to improve the modelling of molecules, such as proteins and drug ligands, within cryo-EM images. Daisuke Kihara, a professor of biological sciences and computer science at Purdue and the leader of Intellicule, emphasises the potential impact. “It will have the potential to accelerate the development of novel drugs by offering precise structural information that can guide the design of molecules with improved efficacy,” Kihara stated.
Intellicule’s software directly addresses a persistent challenge in cryo-EM: difficulty in reliably achieving high-resolution images better than 3 angstroms ($\text{Å}$). When resolution is lower, the manual process of modelling biomolecules and drug molecules becomes time-intensive and susceptible to errors, creating a bottleneck for pharmaceutical companies. The development team, which includes co-founders Charles Christoffer and Genki Terashi, aims to create tools that make the cryo-EM modelling process more accessible and reliable for non-specialists.
Deep learning overcomes resolution hurdles
The Phase I SBIR project centres on utilising cutting-edge deep learning techniques to overcome the current limitations in processing lower-resolution cryo-EM data. Deep learning, a powerful form of artificial intelligence (AI), is particularly adept at image processing tasks.
Kihara noted that deep learning is the core of Intellecule’s solution. “In this software, it enables the detection of atoms in low-resolution cryo-EM images, something that would otherwise be extremely difficult to achieve,” he explained.
By accurately modelling ligand-receptor complexes at resolutions where conventional methods struggle, the software is poised to streamline structure-based drug design.
