Biomarker Identification in Pancreatic Cancer Using Two-Step Feature Selection and WGCNA
Supervisor: Utsha Das, Assistant Professor, Department of CSE, RUET
The thesis develops an integrated computational workflow to investigate potential pancreatic-cancer biomarkers from gene-expression data. It combines statistical differential-expression analysis with feature selection, predictive modeling, co-expression network analysis, PPI-network analysis, and downstream biological interpretation.
Research question
Can complementary feature-selection, predictive-modeling, co-expression, and network-analysis methods reduce high-dimensional pancreatic-cancer gene-expression data into a smaller and biologically meaningful set of candidate biomarkers?
Methodology
My contribution
I designed and executed the integrated analysis workflow, including preprocessing, differential-expression analysis, two-stage feature selection, classifier evaluation, WGCNA, PPI analysis, and downstream biological interpretation.
Final computational candidate panel
NUSAP1, ANLN, MELK, KIF14, NUF2, and TPX2 emerged as the final hub-gene panel investigated in the thesis. These are computational biomarker candidates and require independent and experimental validation before clinical interpretation.