MMS-32: Development and Evaluation of a 32-CpG Salivary DNA Methylation Score Associated with Subjective Memory Outcomes in Males

DNA methylation (DNAm) signatures may capture molecular variation associated with cognitive phenotypes, but saliva-based scores for subjective memory are poorly characterized. We developed MMS-32, a frozen 32-CpG salivary DNAm score derived from an independent EPIC-array comparison of individuals reporting Very bad memory (n = 1,117) versus Very good memory (n = 2,400). Candidate loci were screened using Mann-Whitney U tests; MMS-32 comprises all candidate CpGs with discovery p < 1 × 10−4 and uses signed discovery mean differences as normalized weights. The frozen score was evaluated in an independent male dataset (n = 1,044). Among single-category memory responses, MMS-32 discriminated Very bad (n = 16) from Very good (n = 66) memory with AUC = 0.861 (95% bootstrap CI 0.736–0.957; p = 8.50 × 10−6). A broader Bad/Very bad versus Good/Very good contrast yielded AUC = 0.603 (p = 0.0079). Linear dependence on chronological age was negligible (R2 = 0.0022; Pearson r = −0.047, p = 0.133), and the extreme-memory association remained significant after age adjustment. MMS-32 is therefore a research-stage, memory-associated methylation signature rather than a biological-age score. Independent validation against objective (diagnosed dementia, MRI etc.) and longitudinal cognitive outcomes is required.