Objective: To perform a genome wide association study (GWAS) of stroke recovery phenotypes as defined by the trajectories of patients’ NIH stroke scale scores over time. Background: Stroke recovery research entails tracking multiple domains of neurologic impairment over time that are measured on differing scales. This presents challenges for analytic approaches commonly used to study stroke genetics. Here, we present a novel approach using a network science-based method, Trajectory Profile Clustering (TPC), to define stroke recovery phenotypes for GWAS. Design/Methods: We analyzed data from 3,679 patients in the VISP dataset (Vitamin Intervention for Stroke Prevention) on 15 NIHSS impairment measures at 6 time points spanning 24 months post-stroke. The 4 identified TPC profiles (Figure, Panel A), were used as phenotypes for a subsequent GWAS on the 2,099 genotyped VISP patients. We used a multinomial generalized regression with TPC profiles as the response variable. The model adjusted for age, sex, treatment group, and the first ten principal components. Results: TOPMed imputation and strict quality control resulted in 6,392,745 SNPs. Although none of our loci reached genome wide significance, the GWAS identified several suggestive loci (p<5e-6) clustering within or near genes of biological relevance (Figure, Panel B): LINC02151 (non-coding RNA 11q23.3) is associated with lipid homeostasis, sphingomyelin levels, neurofibrillary tangles and coronary artery disease. SEC63P2 (RNA-seq 4p15.1) is associated with hippocampal volume and is expressed in choroid plexus during development. IDE (10q23.33) is associated with triglyceride levels, Type 2 diabetes and BMI, and in the present analysis also differentiated among the 4 TPC-derived phenotypes (p<0.05). Conclusion: Phenotypes defined using TPC, even in this limited sample, identified multiple suggestive genetic associations with biological relevance for both stroke risk and stroke recovery.