Projects

Omics technologies, encompassing genomics, proteomics, transcriptomics, and metabolomics, have revolutionized our ability to systematically interrogate biological systems, offering unprecedented insights into the intricate dynamics of cells and organisms, revealing the complex interplay of genes, proteins, and metabolites. However, single omics studies often fall short in providing a comprehensive understanding of biological phenomena. To address this limitation, there is a transition into a "multi-omics era," where integrated approaches leverage rapid technological innovations in multiple simultaneous diagnostic tools and high-throughput phenotyping systems.

Despite the potential of multi-omics data to enhance interpretability in results, the analysis and interpretation of such complex datasets pose significant technical challenges. lssues like missing data, low signal-to-noise ratio, and the curse of dimensionality can lead to a prevalence of false positives and false negatives. The initial expectation that collective analysis of multiple omics datasets would mitigate individual shortcomings has not always held true. lntegrating diverse datasets has proven challenging, with many developed methods showing limitations and the absence of a dedicated benchmarking effort complicating method selection even more.

Current challenges in the field include the scarcity of complete multi-omics datasets and the absence of centralized repositories. Few benchmarking platforms include multi-omics datasets, limiting the generalization of methods. The proposed project acknowledges these challenges and critically evaluates integration techniques, underlying approaches, and stages of integration, emphasizing the importance of carefully selecting methods aligned with the study's characteristics rather than adopting a one-size-fits-all approach.

This project aims to tackle these challenges and bridge gaps in current multi-omics integration methods by developing a computational framework that combines data-driven and knowledge-based approaches to address various combinations of multi-omics datasets. Three specific objectives are outlined: (1) creating multi-omics benchmark datasets based on known molecular interactions, (2) constructing a computational framework for benchmarking approaches tailored to various multi-omics subsets, and (3) biologically evaluating the models within the developed framework.

Therefore, the project takes an innovative approach to multi-omics data integration by exploring diverse integration tools and leveraging the diversity across multiple methods designed for specific data combinations and omics levels. Through systematic evaluations and comparisons, the aim is to contribute to the advancement of multi-omics integration methods, offering a balanced and effective solution for diverse experimental settings and data types.

Preliminary data, including knowledge networks such as the Stress Knowledge Map (SKM) and Genome-scale metabolic models (GEMs) developed at NIB, will be utilized in the project. Additionally, relevant datasets have been compiled from international projects. The integration and visualization protocol developed in preliminary stages will be thoroughly refined and utilized in the project's tasks.

In conclusion, this project represents a significant step towards advancing multi-omics integration methods. By addressing critical challenges in the field through innovative approaches, rigorous benchmarking, and adaptation of existing knowledge networks, the project aims to provide a comprehensive solution for diverse experimental settings, opening avenues for new discoveries in the dynamic realm of multi-omics data analysis. 

SICRIS