Early peripheral blood gene expression predicts 90-day outcomes following subarachnoid hemorrhage

Previous clinical, radiological and machine learning studies have predicted 90-day outcomes following subarachnoid hemorrhage (SAH). The present study was designed to determine whether early changes in mRNA expression of immune, clotting and other genes expressed in peripheral blood can predict patient outcomes at 90 days after SAH and possibly provide insights into the molecular factors that promote good versus poor outcomes.

Peripheral blood was drawn after SAH and from vascular risk factor controls (VRFC) and RNAseq performed to measure mRNA expression. A mixed effects regression model identified potential predictors and machine learning algorithms derived the best predictors of 90-day SAH outcome as measured by modified Rankin Score (mRS) for a derivation cohort (23 Poor and 37 Good SAH Outcome patients, 48 VRFC). The model trained on the derivation cohort was then used to predict 90-day SAH outcome in an independent validation cohort (15 Poor and 23 Good SAH Outcome). Enrichment analyses for cell-type specific genes, canonical pathways, and biological processes were performed for the predictor genes.

The mixed effects regression on the derivation cohort yielded 94 genes from which 20 were selected through feature reduction. Machine learning algorithms were optimized to generate a model that predicted SAH 90-day outcome with AUC = 0.85, sensitivity = 87%, and specificity = 84% on cross-validation. Application of this model to the independent validation cohort yielded AUC = 0.84, sensitivity = 93%, and specificity = 74%. The 20 predictors were significantly enriched in genes from neutrophils and erythroblasts and in nine pathways including the Unfolded Protein Response, Neutrophil Degranulation, and Neutrophil Extracellular Trap Signaling.

This discovery study demonstrates that a small panel of 20 genes expressed in peripheral blood after SAH has the potential for predicting 90-day outcomes following SAH. It also shows that neutrophils may be important drivers of SAH outcomes and could represent therapeutic targets.