Data source

The gene expression profile of the prefrontal cortex brain tissues of 310 AD patients and 157 non-demented control samples were retrieved from the {"type":"entrez-geo","attrs":{"text":"GSE33000","term_id":"33000"}}GSE33000 dataset⁸ of the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database. This dataset was selected. The processed data, which have been adjusted for the age, gender, RIN, pH, PMI, batch, and preservation of the samples, were downloaded from the Sample table. This dataset contains 39,279 detected probes, of which 13,798 were annotated, and a total of 9969 genes were profiled, while 31 probes were omitted due to mapping to more than one gene.

Another publically available microarray dataset {"type":"entrez-geo","attrs":{"text":"GSE84422","term_id":"84422"}}GSE84422⁹, which profiled PFC from 56 postmortems with varying degrees of AD pathological abnormalities, was utilized as the unseen dataset to verify our models. The samples were classified into control or AD by CDR, Braak, and CERAD. Notably, due to the difference of microarray used, out of the 9966 attribute genes of the training dataset, 3680 genes were not profiled in the testing dataset. To conduct the testing, these 3680 gene profiles were artificially added with FC assigned as "1" for all samples.