GIDAS: gene expression analysis across two diseases, ending in a co-authored research abstract

I. Stakes

Public gene-expression data is already collected, already normalized, and free to download, so the bottleneck is not the biology but whether you can run a clean pipeline over it and read the result honestly. GIDAS runs one differential-expression workflow twice, first on tuberculosis and then on Alzheimer's disease, on datasets pulled from the Gene Expression Omnibus. Build the pipeline once, apply it twice, and see how far the signal carries.

II. Method

One pipeline, pointed twice

Tuberculosis came first. I pulled GSE19444, the active-TB whole-blood microarray cohort, ran a standard limma-style differential expression pipeline, and ranked probes and genes by log fold-change and adjusted p-value. The top up-regulated and down-regulated hits then get imported into STRING, which maps them onto known protein-protein interactions so you can see which network modules cluster around the differential signal instead of reading a flat gene list.

The same workflow, aimed at Alzheimer's

The Alzheimer's work reused the pipeline against the relevant GEO cohorts, this time focused on the NMDA receptor signaling pathway. Pull the cohort, run differential expression, intersect the hits with curated KEGG and Reactome pathways, then run STRING enrichment to see which interaction modules cluster around the candidate genes. The analysis surfaced upregulated phosphorylation as a candidate contributor to long-term potentiation impairment in early-stage disease. Pairing the cross-cohort signal with pathway enrichment was the contribution.

III. What shipped

The Alzheimer's analysis is listed as a contributing-author credit on a paper titled “Upregulated phosphorylation in Alzheimer's disease, NMDA receptors and LTP.” The named contribution is specific: differential expression analysis, STRING network construction, and figure generation. That is the honest scope. The computational half rather than the wet-lab biology behind the claim, the part that turns public data into a ranked, enriched result the paper reports.

Bioinformatics needs high-school-level statistics and real biology, and not much else. With the right scripts and the right datasets, a student can produce signal that ends up in a research publication.

The gap worth naming: the paper's journal, publication date, and DOI are not recorded here, so the credit is documented by the contribution rather than a live citation. The verifiable half is the work itself, the accession numbers, the ranked CSVs, and the STRING subnetworks, all of which exist and reproduce.

∎