Establishing edge labels for benchmarking gene co-expression networks of saccharomyces cerevisiae and homo sapiens

• 主要作者: Antony Velankanni Jenet Princy
• 其他作者: Marek Mutwil
• 格式: Final Year Project
• 語言: English
• 出版: Nanyang Technological University 2024
• 主題: Medicine, Health and Life Sciences
• 在線閱讀: https://hdl.handle.net/10356/176362

The reliance on one type of correlation to construct Gene Coexpression Networks (GCNs) using ensemble approaches hinders a comprehensive understanding of gene co-expressions. To address this, Lim et al. (unpublished) propose a Two-Tier Ensemble Aggregation (TEA) GCN, combining various GCNs generated using different correlation coefficients and different dataset partitions for Arabidopsis thaliana. This project extends the evaluation of TEA-GCN to Saccharomyces cerevisiae and Homo sapiens through edge labelling and benchmarking TEA-GCN against other state-of-the-art ensemble methodologies. To this end, we used gene information from Metabolic pathways, Gene Ontology, and Transcription factors by utilising gene lists from public databases to generate labelled edges. These edges were further refined to obtain the best evaluation data for network performance assessment using Receiver Operating Characteristic and Precision-Recall curves. We show that TEA-GCN outperformed the state-of-the-art GCNs for both the species, in addition to Arabidopsis thaliana. This has demonstrated that the robustness of TEA-GCN’s methodology is generalisable across the diverse transcriptional programs underpinning not just fungal, plant, and mammalian species, but also single-cellular and multicellular organisms. Furthermore, this endeavour has resulted in first-ever comprehensive benchmarking data for Homo sapiens and Saccharomyces cerevisiae GCNs which will be instrumental to the development of more advanced GCN methods.