AI and machine learning (ML) are revolutionizing computational biology and Biolexis Therapeutics, a preclinical-stage biopharmaceutical company, is at the forefront of this revolution. Biolexis is using AI/ML to develop small-molecule therapeutics targeting a range of diseases for precision medicine.

In just the last few years, research has demonstrated that AI-driven applications can generate accurate protein structure models , providing a deeper understanding of a protein’s function and its effect on cellular structures and functions. Where once research into protein structure and function for drug development took years of trial and error, researchers now leverage AI to do it in a fraction of the time with a substantial increase in accuracy. 

Biolexis developed FIELDS™, an advanced AI/ML drug discovery platform utilizing laboratory-generated molecular data in conjunction with deep learning for model training and testing of protein structures. This facilitates accurately predicted structural models by mapping hot spots and computing ligand properties to discover active, safe, and effective Novel Chemical Entities (NCEs) for precision medicines. Additionally, FIELDS™ reduces therapeutic development costs and accelerates timelines while increasing the probability of success in First-In-Human (FIH) clinical trials. 

LinkThe Complexity of Biochemical Datasets

Biolexis’ proprietary database of 500,000 biochemical complex data points, as well as public biology and chemistry datasets amounting to billions of records, posed a significant AI/ML challenge. Compounding the problem is that some of the tools used for molecular dynamics and docking were originally created to run on a single machine or research supercomputer. It’s possible to use them on a cluster but, ultimately, it was difficult to process and manage data movement.  Cloud optimization wasn’t even considered when these tools were originally developed. As a result when utilizing these prior tools, just one of the Python programs would take a full seven days to process a workload. 

LinkThe Anyscale Platform and Ray - Faster, Scalable Outcomes

Ray and the Anyscale Platform enabled Biolexis to go from zero to a POC and fully scaled in production with little or no friction.  The Anyscale Platform also made using the same toolset a cloud native experience. Anyscale Workspaces, a unified seamless developer experience, ensures environments are workload ready and scalable to the needs of internal scientific teams and customers.

“Anyscale Workspaces enables us to use development tools such as Jupyter notebooks that we are already familiar with, while allowing us to seamlessly scale our workloads. Anyscale Workspaces allows us to also experiment at scale, increase the number of iterations and experiments, and accelerate AI-driven scientific outcomes.” Evan Marshman, Data/ML Engineer at Biolexis

In just the POC phase the Anyscale Platform enabled workload scaling and accelerated the same workload that originally took seven days to process, to under five hours. That’s approximately a 56x improvement.

“We’ve already experienced game changing improvements after adopting the Anyscale Platform and Ray and have effectively launched the FIELDS platform, which uses model training and test sets to identify novel fragments of small molecules for any given (un)druggable protein, RNA-based, PROTAC, AUTOTAC, LYTAC and Molecular Glue target of interest , for general use.” Hariprasad Vankayalapati M. Pharm, PhD CSO Biolexis Therapeutics.

With the Anyscale Platform, Biolexis can now parallelize almost any process and scale it easily.  Scaling is only constrained by availability of AWS infrastructure.. Additionally, the Anyscale Platform, which provides Ray and additional capabilities as a fully-managed service, simplifies moving workloads to production at scale without requiring a dedicated infrastructure team.  The advantages of the Anyscale Platform and Ray allow Biolexis to maintain a small technology team while delivering AI outcomes quickly, which ultimately provide our customers—clinical agents—faster scientific outcomes at a lower cost.

“We are excited to continue advancing our AI/ML-augmented discovery platform with the Anyscale Platform and Ray.”

David Bearss PhD CEO Biolexis Therapeutics