How multiomics and AI advance drug discovery

Human biology is a dynamic and complex system. As a result, human diseases are inherently interconnected, which poses a significant challenge for drug discovery and treatment. Traditional research has typically focused on a single disease, often overlooking the broader context of human health. However, an exciting development in recent years – multiomics – is changing things for the better.

This combined study of different ‘omic’ fields has been identified as crucial for advancing drug discovery and accelerating the success of candidates sent to clinical trials. By integrating all ‘omic’ disciplines, scientists are able to gain a more comprehensive and accurate understanding of biology as a whole. Rather than focusing on a single layer, such as genomics, multiomics enables researchers to gain a fuller picture of the dynamic interactions at play in biological organisms. By combining multiomic data and AI, companies like BPGbio are able to better model diseases and enhance the drug discovery process.

Despite the growing interest in AI and machine learning, many AI-driven therapeutic approaches have failed to meet expectations, with a high failure rate in clinical trials and regulatory approvals. BPGbio stands out as a notable exception, with several assets currently in phase 2b trials for aggressive cancers and no failed clinical trials to date. Its success can be attributed to the integration of clinically annotated data with full multiomic datasets, feeding its Bayesian/causal AI model, the NAi Interrogative Biology Platform.

This approach combines systems biology, integrative biology, and cutting-edge AI to uncover causal relationships in disease biology, which is especially critical for understanding co-morbidities and the intricate interactions between inflammation, metabolic disease, and neurological disease.

A comprehensive multiomic approach powered by AI

By focusing on multiomics, BPGbio is able to capture the complete and nuanced view of a patient’s biology. "We want to treat the patient, not just the disease,” explains Michael Kiebish, Head of Platform and Translational Sciences at BPGbio. This patient-centric mindset helps the team identify drug candidates which consider the interconnectivity of disease, while understanding how comorbidities affect disease progression and drug responses in a broad range of patient types –overall improving the likelihood of identifying clinically relevant drug targets.

Through the integration of a wide array of omic technologies, BPGbio is able to generate a deep data set enabling accurate predictions about potential drug targets and clinical outcomes. As Kiebishgoes on to explain, “we base our work on causation, not correlation.” By starting with a comprehensive approach, the team can identify novel, actionable targets which may have been overlooked by other approaches.

Unlike other AI platforms that rely on correlation-based analysis, BGPbio employs Causal AI methodologies. Its NAi Interrogative Biology Platform uses Bayesian causal inference techniques to enable the discovery of new targets without bias from existing data. Rather than being constrained by prior knowledge, the team allows the data to drive discoveries.

"We want to actually discover targets from the data alone within no bias, and we want to let the data actually speak for itself," comments Kiebish, an approach which minimizes the risk of pursuing a hypothesis which may be based on faulty or incomplete data – a problem keenly associated with the traditional drug discovery process.

In-house expertise for high-quality data

While many companies choose to outsource its wet lab work, BPGbio takes a different approach. With an in-house wet lab, they maintain complete control over the process, ensuring the highest quality and consistency in the data generated. This also enables the rapid validation of AI-generated predictions and refining of models using real-world experimental data. By combining AI-driven insights with biological validation, BPGbio can confidently advance its candidates to clinical trials.

Collaborations to propel the future

While BPGbio has extensive experience in omic technologies, the company also recognizes the benefits that collaborations have on the success of its projects. "We’ve strived to always maintain that level of robustness and be part of ongoing discussions in the field on how to make approaches better and how to utilize technologies effectively," Kiebish explains. Through this work, the company has built an expanding network of strategic partners, including Mount Sinai School of Medicine and Oxford.

Critically, there is an ongoing collaboration with the U.S. Department of Energy, which provides BPGbio with access to Frontier, the world’s fastest supercomputer. AI applications face numerous challenges, including high energy consumption and the need for standardized, reproducible omic data. At the moment, this collaboration ensures they are able to leverage cutting-edge technology and handle the vast amounts of data generated by its Bayesian AI models. Looking ahead, the team aims for enhanced computational efficiency to boost the throughput of these models.

“The entire industry's challenge is standardization in reproducibility,” states Kiebish.“We've come a long way in the past 10 years but we still have a long way to go, even with AI, and making it computationally efficient is still a major challenge across the board.”

Looking ahead, Kiebish sees spatial omics playing a pivotal role: "Spatial omics is where it will evolve and where AI will have the best success, because it’s a society of cells that are communicating. You can actually see the data and seeing is believing rather than making a bunch of bar graphs."

There is no doubt that BPGbio’s innovative approach is setting a new standard for clinical omics and personalized medicine across the board, offering an exciting path forward to more effective treatments for cancers and other complex diseases.