Challenges in Integrating Transcriptomics and Metabolomics Data in Cancer Research

Author(s): Maryna Chepeleva,Petr V. Nazarov

Affiliation(s): Luxembourg Institute of Health



Understanding the molecular mechanisms driving cancer is crucial for developing of effective treatment strategies. Transcriptomics uncovers key regulatory pathways and molecular signatures associated with disease progression, while metabolomics provides a snapshot of cancer metabolic phenotype, highlighting metabolic reprogramming and dysregulated pathways driving cancer growth and survival. However, the direct linkage of transcriptomic and metabolomic levels of regulation is still challenging but offers a promising direction for unraveling the complexities of cancer processes. Here we are working on developing strategies for linking metabolite profiles with gene expression patterns based on matrix factorization. We analyzed several datasets, including melanoma NRAS-mutant cell lines and glioblastoma samples, which contain both transcriptomics and metabolomics data levels. Applying our previously developed and published Bioconductor R package consICA, we aim to decompose transcriptomic data and extract hidden molecular signals. The package not only implements robust consensus independent component analysis but also annotates components and provides various approaches for their analysis. Integrating datasets, we investigated the changes in two modalities and identified potential cross-talk mechanisms. By selecting the most discriminative features, we tried to enhance the predictive accuracy and interpretability of our models. Additionally, we observed the limitations in linking transcriptomics to metabolomics using flux analysis and characterized the robustness of published models such as scFEA.