AI on OMICS Data

AI on OMICS Data

Omics data refers to large-scale biological datasets, such as genomics, transcriptomics, and proteomics, that provide a comprehensive view of biological systems. AI is used to process and analyse these complex datasets, allowing for the identification of patterns and relationships that can lead to new insights into biological processes and disease mechanisms. Within IDLab Ghent, we are focusing on:

• Identification of molecular biomarkers for medical decision making in oncology
• Improved drug target prioritization using genetic and/or phenotypic screening data that allow mode-of-action identification
• Identification of splice variants from single cell data: Splice variants are different versions of a gene that can have different functions and identifying them from single cell data can help researchers understand what goes wrong in diseases.

At IDLAB Ghent, we pride ourselves on the incorporation of state-of-the-art data mining techniques with the proper biological assumptions. This enables us to speak the language of industry. For example, we use graph-based analysis techniques to integrate omics data with prior information on molecular interactions (probabilistic approaches, random walk-based approaches, link prediction) and deep learning techniques for image analysis. Additionally, the data is often somewhat noisy and can be large-scale.

Within IDLab, we apply OMICS data in two major fields:

• Human health, where we focus on cancer research
• Plant and animal breeding applications

Our graph-based approaches are widely applicable in bioinformatics, as is our work on DNA/RNA sequencing data analysis. We often seek to use multi-modal approaches, combining, e.g., image data with genomics, to capture phenotypical as well as genotypical information. This also allows linking a phenotypical expression of disease to genetic drivers.