HTLV-1-associated myelopathy (HAM/TSP) is a chronic and progressive inflammatory disease of the central nervous system. The aim of our study was to identify genetic determinants related to the onset of HAM/TSP in the Japanese population. We conducted a genome-wide association study comprising 753 HAM/TSP patients and 899 asymptomatic HTLV-1 carriers. We also performed comprehensive genotyping of HLA-A, -B, -C, -DPB1, -DQB1, and -DRB1 genes using next-generation sequencing technology for 651 HAM/TSP patients and 804 carriers. A strong association was observed in HLA class I (P = 1.54 × 10-9) and class II (P = 1.21 × 10-8) loci with HAM/TSP. Association analysis using HLA genotyping results showed that HLA-C*07:02 (P = 2.61 × 10-5), HLA-B*07:02 (P = 4.97 × 10-10), HLA-DRB1*01:01 (P = 1.15 × 10-9) and HLA-DQB1*05:01 (P = 2.30 × 10-9) were associated with disease risk, while HLA-B*40:06 (P = 3.03 × 10-5), HLA-DRB1*15:01 (P = 1.06 × 10-5) and HLA-DQB1*06:02 (P = 1.78 × 10-6) worked protectively. Logistic regression analysis identified amino acid position 7 in the G-BETA domain of HLA-DRB1 as strongly associated with HAM/TSP (P = 9.52 × 10-10); individuals homozygous for leucine had an associated increased risk of HAM/TSP (odds ratio, 9.57), and proline was protective (odds ratio, 0.65). Both associations were independent of the known risk associated with proviral load. DRB1-GB-7-Leu was not significantly associated with proviral load. We have identified DRB1-GB-7-Leu as a genetic risk factor for HAM/TSP development independent of proviral load. This suggests that the amino acid residue may serve as a specific marker to identify the risk of HAM/TSP even without knowledge of proviral load. In light of its allele frequency worldwide, this biomarker will likely prove useful in HTLV-1 endemic areas across the globe.