Panel-Gene Transcriptomic Associations in Amyotrophic Lateral Sclerosis: a donor-resolved seven-cohort analysis of a locked 468-gene detoxification panel

Abstract Amyotrophic lateral sclerosis (ALS) was analysed in seven adult central-nervous-system cohorts (GSE153960, GSE26927, GSE76220, GSE19332, GSE68605, GSE56500 and GSE18920; 220 ALS and 97 control donors) with a locked 468-gene redox, metal, xenobiotic and barrier panel. Each cohort entered as one estimate per donor: bulk spinal cord in GSE153960, with cervical, lumbar and thoracic counts summed per donor and sequencing platform modelled; donor averages in GSE26927; motor-neuron samples only in GSE18920; and gene-level summaries for the two exon-array cohorts. Random-effects meta-analysis yielded 449 analysable genes, 33 at false-discovery rate (FDR) < 0.05 and 28 high-confidence genes (FDR < 0.05, measured in ≥ 3 cohorts, direction-concordant in ≥ 70% of cohorts, nominally significant in ≥ 2 cohorts); 25 of the 28 were higher in ALS, and the median I² among them was 56.7%. Eight genes — ABCA1, APOE, AQP1, AQP4, SLC30A7, SLC40A1, SLC7A11 and SP1 — remained high-confidence in every sensitivity analysis of the primary design, including adjustment for neuronal, myeloid and astrocytic marker scores, which reduced the high-confidence set to 11–14 genes; six of the eight did so when the largest cohort, GSE153960, was left out. Against coverage-matched random gene sets, the panel was not over-represented among FDR-significant (35 v 41.5 expected) or high-confidence genes (29 v 26), and the Hartung–Knapp–Sidik–Jonkman adjustment left one gene. These results revise an earlier locked analysis of the same seven cohorts (456 analysable, 26 FDR-significant and 22 high-confidence genes; median I² 60.6%) whose inputs pooled all tissues of GSE153960 and both tissue compartments of GSE18920 and counted GSE26927 replicate arrays as separate subjects. Every number is transcript abundance in postmortem tissue.