Expert Curation of Human Microarray Probe Annotations in the Biological Interpretation of Integrated Transcriptome Analyses

Background: Accurate mapping of microarray probes to gene loci is essential for a reliable interpretation of gene expression data, yet available probe annotations are often incomplete or outdated. In this study, we aimed to develop and apply an expert-curated approach for the reannotation of human microarray probes and to assess its impact on transcriptome data interpretation. Methods: Human microarray probe sequences were reannotated by integrating information from platform databases with BLAST alignment against updated reference transcripts, followed by manual expert curation of probe-to-locus assignments. As a proof of concept, we reanalysed an integrated normal-human-heart transcriptome map, focusing on 17 genes with high and 17 with low expression variability. We then applied the curated annotation to 1032 probes representing genes involved in one-carbon metabolism in an integrated transcriptome map of trisomy 21 versus euploid human fibroblasts. The impact of reannotation on gene expression estimates was assessed, and eight selected genes were experimentally validated by Real-Time RT-PCR in independent fibroblast samples. Results: Manual curation identified annotation discrepancies for 132/357 (37.0%) probes among genes with high expression variability and 24/43 (55.8%) probes among genes with low expression variability in the heart transcriptome map. In the trisomy 21 versus euploid fibroblast dataset, 341/1032 (33.0%) probes required reassignment or removal. Incorporation of the curated annotations modified expression estimates and improved their consistency for multiple genes. Experimental validation of eight selected genes showed a positive correlation between Real-Time RT-PCR and reannotated transcriptome data (r = 0.67). Conclusions: Expert-curated reannotation of microarray probes identified substantial annotation discrepancies and corrected probe-to-locus assignments affecting a considerable proportion of probes, with measurable consequences for gene expression estimates. The resulting set of 1432 manually curated probe annotations provides an updated resource for the interpretation of legacy human microarray datasets and highlights the need for more comprehensive and usable automated pipelines capable of reproducing expert-level probe assignment.