Computational Molecular Biology 2024, Vol.14, No.2, 84-94 http://bioscipublisher.com/index.php/cmb 85 2 Overview of Biological Network Analysis 2.1 Types of biological networks Biological networks are used to represent various interactions within a living organism, encompassing multiple types of relationships. Gene Regulatory Networks (GRNs), which model how genes interact to control gene expression, are central to understanding cellular processes. Protein-Protein Interaction (PPI) Networks track physical interactions between proteins, revealing key connections in signal transduction and cellular function. Metabolic Networks illustrate the biochemical reactions within cells, showing how different molecules are converted by enzymes to maintain life (Schmitt et al., 2023). Signaling Networks describe the pathways through which cells respond to external stimuli, and Phenotype Networks link gene variants to observable traits. Each network type provides insights into different aspects of biology, from cellular metabolism to complex traits and diseases. Recent technological advances in high-throughput omics, including transcriptomics, proteomics, and metabolomics, allow for large-scale biological data collection. This data has been pivotal in reconstructing these networks to study diseases such as cancer and Alzheimer's disease (Shojaee & Huang, 2023; Hill et al., 2016). However, the high-dimensionality and dynamic nature of biological networks make their analysis complex, requiring sophisticated computational methods to model and understand their intricacies (Furqan & Siyal, 2016; Buetti-Dinh et al., 2020). 2.2 Commonly used network analysis techniques Several techniques are commonly used in the analysis of biological networks. Graph-theoretical approaches form the backbone of most network analyses, where nodes represent genes, proteins, or metabolites, and edges represent interactions between them. These methods are powerful for identifying key nodes (hubs) or pathways that are central to network integrity. Clustering algorithms, such as hierarchical clustering and k-means, group similar nodes to find functional modules within networks. Bayesian Networks and Markov Models offer probabilistic frameworks for inferring causal relationships between components, which are especially useful in gene regulatory and metabolic network analyses. Granger Causality is another method frequently applied to infer cause-effect relationships in time-series data, particularly useful in dynamic biological processes like gene expression regulation. Additionally, machine learning techniques, such as deep learning and graph neural networks, have gained traction in recent years due to their ability to handle large and complex datasets. These methods can predict regulatory relationships, allowing for more accurate network reconstructions, especially in large-scale transcriptomic and proteomic studies (Furqan & Siyal, 2016; Monneret et al., 2017). However, each technique has limitations that need to be addressed depending on the type of biological data and the specific questions being asked. 2.3 Limitations of correlation-based approaches Correlation-based approaches, while commonly used in biological network analysis, have significant limitations in inferring causal relationships. These methods primarily capture associations between variables, which do not necessarily imply a direct cause-and-effect relationship. For instance, two genes may exhibit a high correlation in expression levels, but this may be driven by a common upstream regulator rather than a direct interaction. Furthermore, correlation-based methods cannot account for confounding variables, leading to false positives or misinterpretations. These approaches are also limited in their ability to handle non-linear relationships and dynamic changes over time, which are critical features of biological networks. For example, gene expression patterns can be highly context-dependent, changing in response to environmental stimuli or developmental stages, which simple correlations fail to capture. 3 Fundamentals of Causal Inference 3.1 Definition and concepts of causality Causality, in the context of biological networks, refers to the influence one biological component (such as a gene or protein) exerts over another. A causal relationship implies that changes in one entity (the cause) directly lead to changes in another (the effect), often under experimental or observational conditions. This concept is crucial in biological research for understanding complex pathways, disease mechanisms, and potential therapeutic
RkJQdWJsaXNoZXIy MjQ4ODYzNA==