Other Feb 2023 – Sep 2023

Understanding the Application of Machine Learning in Cell Signalling Pathway Analysis and Developing a DEG Analysis Handbook

Author / Contributor
Maryam Rezaee (Undergraduate Researcher)
Affiliation / Advisor Dr. Fatemeh Mansoori, Department of Mathematics, University of Isfahan, Iran
Venue / Context Bachelor's Research Project Graded 20/20 (Perfect), Department of Mathematics, University of Isfahan, Iran
Keywords: Bioinformatics Machine Learning Differential Gene Expression Cell Signalling Computational Biology

Overview

High-throughput transcriptomic sequencing generates vast quantities of high-dimensional genomic data, presenting significant analytical challenges for computational biology. Identifying disease biomarkers and deciphering dysfunctional cellular communication requires robust statistical models and machine learning pipelines capable of isolating signal from biological noise.

This undergraduate research project, conducted under the supervision of Dr. Fatemeh Mansoori at the University of Isfahan, explores the application of classical and modern machine learning algorithms to cellular signaling pathway analysis. The project culminated in two primary deliverables: a systematic literature review of machine learning formulations across bioinformatics domains, and an implementation handbook designed to guide future researchers through Differentially Expressed Gene (DEG) pipelines.

Machine Learning in Pathway Analysis

Biological signaling pathways operate as complex directed networks where coordinated changes in gene expression regulate cellular state and response. Traditional univariate statistical tests often fail to capture multivariate gene interactions and non-linear regulatory dependencies.

Our review examined supervised and unsupervised learning frameworks applied to transcriptomic data, focusing on Support Vector Machines (SVMs), regularized linear models, and dimensionality reduction techniques. We analyzed how kernel methods project high-dimensional gene expression vectors into separable feature spaces, enabling effective patient classification and subgroup stratification. The investigation evaluated methods for mapping identified gene subsets back onto canonical signaling topologies, identifying functional pathway enrichment and regulatory bottlenecks associated with pathological states.

The DEG Analysis Handbook

A core outcome of this project was the authorship of a comprehensive technical handbook and implementation roadmap for Differentially Expressed Gene analysis. Navigating transcriptomic pipelines requires addressing multiple sequential stages:

  1. Quality Control and Preprocessing: Raw count filtering, sequence alignment verification, and batch-effect detection.
  2. Normalization and Variance Stabilization: Evaluating log-ratio transformation, Trimmed Mean of M-values (TMM), and variance-stabilizing transformations to account for library size differences.
  3. Statistical Modeling: Fitting generalized linear models (GLMs) based on negative binomial distributions to handle overdispersed count data.
  4. Multiple Testing Correction: Implementing False Discovery Rate (FDR) control via Benjamini-Hochberg procedures to prevent type I error inflation across thousands of parallel gene hypotheses.
  5. Pathway Enrichment and Visualization: Volcano plot generation, hierarchical clustering, and Gene Ontology (GO) enrichment mapping.

The resulting handbook provides clear conceptual explanations, reproducible code workflows, and diagnostic guidelines, serving as an onboarding curriculum and practical reference for students entering computational genomics.

Uplink Ready

Lab Dispatch Uplink Reserved

Technical writeups, experiment traces, and theoretical musings for this work will synchronize here as the lab.maryamrezaee.me ecosystem comes online.