The Critical Role of Biological Data in Advancing AI Drug Discovery

GSK and Relation Therapeutics Collaborate in AI-Assisted Drug Discovery

GSK has partnered with British biotechnology company Relation Therapeutics in a research collaboration valued at up to $110 million. This collaboration expands on their existing work in AI-assisted drug discovery.

Relation will be responsible for generating large-scale datasets to measure how human cells respond to genetic changes and drug interventions. These datasets will be utilized to train AI models that aim to identify potential drug targets, including those within Relation’s MORGAN platform.

The agreement emphasizes the integration of biological data generation with AI model development. Relation’s research methodology combines computational analysis with experimental data generation on human cells.

This collaboration builds upon previous agreements between GSK and Relation that focused on fibrotic diseases and osteoarthritis. These projects involved observational studies to create functional disease datasets for analysis using Relation’s Lab-in-the-Loop platform.

Prior work involved a combination of human genetics, single-cell multi-omics data from human tissue, functional assays, and machine learning to identify and validate potential disease targets.

Relation’s Approach to Biological Data Generation

Relation employs a Lab-in-the-Loop approach, which combines laboratory experiments and computational analysis. This includes tissue profiling, single-cell and spatial transcriptomics, sequencing, target validation, and the use of machine learning for target identification, prioritization, validation, and experimental design.

The company also conducts perturbation experiments to understand how genetic changes impact cellular characteristics associated with disease. Results from these experiments are analyzed alongside genetic and patient-derived biological data.

While public repositories such as CZ CELLxGENE, the Human Cell Atlas, and NCBI Gene Expression Omnibus offer vast amounts of single-cell data, combining information from various studies can pose technical challenges.

Assembling a high-quality, non-redundant dataset is crucial for building robust single-cell foundation models, as highlighted in a review published in Experimental & Molecular Medicine.

Challenges in Building Biological Models

Recent research in Nature Methods explored the impact of dataset size and diversity on single-cell foundation models. The study revealed that increasing training data did not consistently improve model performance, stressing the need to balance model capacity, dataset size, and computational resources.

Another study published in Genome Biology evaluated the performance of Geneformer and scGPT models in zero-shot evaluation tasks, highlighting challenges related to batch effects and the assumption that larger models always yield better biological representations.

Pharma Companies’ Focus on Specialized Datasets

Relation has applied its data-generation approach to Osteomics, a functional single-cell bone atlas project that investigates disease biology, therapeutic targets, biomarkers, and patient subgroups in osteoporosis using patient-derived samples.

Specialized dataset providers are becoming increasingly important in AI drug discovery, as seen in agreements like GSK’s collaboration with Ochre Bio for human liver single-cell and perfused-organ data licensing.

Access to high-quality data remains a challenge in AI drug discovery, with companies pursuing various approaches to obtain data and computational capabilities, ranging from AI platform access to joint development and data licensing.

The GSK-Relation collaboration encompasses both data generation and model development, with Relation producing human cellular datasets for training AI models in identifying potential drug targets.

For more information on AI and big data, explore the AI & Big Data Expo happening in Amsterdam, California, and London, co-located with other leading technology events as part of TechEx.

AI News is powered by TechForge Media. Discover upcoming enterprise technology events and webinars here.