Molecular Complexity of Multiple Sclerosis Through Pathway Clustering

Pathway clustering methods are instrumental in elucidating the complex molecular underpinnings of diseases like Multiple Sclerosis (MS). These techniques provide a framework for understanding the biological pathways and molecular mechanisms involved in disease progression and response to treatment. Here, we delve into the intricacies of pathway clustering methods, highlighting common mistakes, crucial parameters, and specific methods such as Over-Representation Analysis (ORA), Gene Set Enrichment Analysis (GSEA), single sample GSEA (ssGSEA), and variant-based pathway clustering. We also explore their applications in complex diseases like MS, citing a study by Avsar et al. (2015) which investigates the etiopathogenesis of MS clinical subtypes using proteomic and bioinformatics approaches in cerebrospinal fluids.

Pathway Clustering Methods in Complex Diseases

Pathway clustering involves grouping biological pathways that exhibit similar patterns in terms of gene expression, mutations, or other molecular data. This approach aids in understanding the systemic organization of biological processes and how alterations in these processes can lead to complex diseases.

Key Parameters

Data Quality and Preprocessing: It's imperative to use high-quality, well-normalized data. Failing to properly preprocess data can lead to inaccurate clusters and misleading biological interpretations.

Algorithm Selection: Choosing the right clustering algorithm is critical. Algorithms like hierarchical clustering, k-means, or consensus clustering each have their strengths and should be selected based on the dataset and the specific analytical needs.

Feature Selection: Including irrelevant features can lead to noisy and uninterpretable clusters. It's essential to select features that are biologically relevant and contribute to the variability you are interested in.

Validation: Clustering results should always be validated, either statistically, by comparing with known labels, or by assessing the biological relevance of the clusters.

Overfitting and Generalizability: Models that are too complex for the amount of data can overfit, leading to poor generalization to new data.

Specific Pathway Clustering Methods

Over-Representation Analysis (ORA): ORA involves checking if a predefined set of genes (like a pathway) is over-represented in a list of genes of interest. It's a straightforward method but can miss signals in genes not included in the predefined sets.

Gene Set Enrichment Analysis (GSEA): GSEA ranks all genes based on their correlation with a phenotype and tests if predefined gene sets (pathways) are enriched at the top or bottom of this ranked list. This method considers all genes, not just those above a specific cut-off, providing a more comprehensive analysis.

Single Sample GSEA (ssGSEA): An extension of GSEA, ssGSEA allows the analysis of pathway enrichment in individual samples, making it possible to understand variations within a disease subtype or between different individuals.

Variant-Based Pathway Clustering: This method focuses on clustering pathways based on the presence of genetic variants. It's particularly useful in genetically complex diseases like MS, where multiple genes and pathways contribute to the disease phenotype.

Application in Multiple Sclerosis (MS)

Our groups study by Avsar et al. (2015) is a prime example of using pathway clustering methods to understand complex diseases like MS. The authors aimed to investigate the molecular pathways involved in different clinical subtypes of MS using proteomic and bioinformatics approaches. They analyzed the CSF proteomic profile from a cohort of 179 patients with different MS subtypes and identified common disease pathways shared by all disease subtypes, such as the renin-angiotensin system and the complement and coagulation cascade pathway. They also identified clinical subtype-specific pathways, providing insights into the pathology and clinical heterogeneity of MS​​.

Another study of our group by Everest et al. (2023), "Prospective outcome analysis of multiple sclerosis cases reveals candidate prognostic cerebrospinal fluid markers", the authors sought to identify predictive biomarkers and understand the long-term disability outcomes in Multiple Sclerosis (MS) patients. This comprehensive study harnessed the power of machine learning and proteomics to analyze cerebrospinal fluid (CSF) data collected from MS patients over a significant follow-up period. Specifically, the initial proteomics data were re-analyzed using a machine learning-based genetic algorithm to identify potential prognostic biomarkers for poor prognosis after a substantial follow-up period. The study also scrutinized clinical and MRI characteristics at disease onset that correlated with long-term disability outcomes. Notably, CSF levels of alpha-2-macroglobulin, apo-A1, haptoglobin, and specific cerebral lesion loads were found to be significantly higher in the group with an unfavorable course of the disease, suggesting their potential role as prognostic markers. The study is groundbreaking in its approach, combining advanced computational methods with clinical data to uncover valuable insights into the prognosis of MS​​​​.

Furthermore, our group's another study used hierarchical clustering of enriched pathways to establish representative pathways for complex diseases, including MS. The clustering approach involved calculating overlap indices between enriched pathways and then performing hierarchical clustering based on these indices. This method enabled the identification of biologically relevant clusters and representative pathways, offering a comprehensive understanding of the disease-associated molecular mechanisms​​.

Conclusion

Pathway clustering methods offer powerful tools for unraveling the molecular complexity of diseases like MS. By understanding common pitfalls, selecting appropriate parameters, and applying specialized methods like ORA, GSEA, ssGSEA, and variant-based clustering, researchers can gain profound insights into the molecular underpinnings of diseases. Studies like the one conducted by Avsar et al. (2015) exemplify the potential of these methods to provide a deeper understanding of the heterogeneity and complex pathophysiology underlying diseases like MS, paving the way for personalized medicine and targeted therapies.

Reference:

Avsar, T., Durası, İ. M., Uygunoğlu, U., Tütüncü, M., Demirci, N. O., Saip, S., ... & Tahir Turanlı, E. (2015). CSF proteomics identifies specific and shared pathways for multiple sclerosis clinical subtypes. PloS one, 10(5), e0122045.
Everest, E., Uygunoglu, U., Tutuncu, M., Bulbul, A., Onat, U. I., Unal, M., ... & Siva, A. (2023). Prospective outcome analysis of multiple sclerosis cases reveals candidate prognostic cerebrospinal fluid markers. Plos one, 18(6), e0287463.
Everest, E., Ülgen, E., Uygunoglu, U., Tutuncu, M., Saip, S., Sezerman, O. U., ... & Turanli, E. T. (2021). Investigation of multiple sclerosis-related pathways through the integration of genomic and proteomic data. PeerJ, 9, e11922.