Large-scale DNA sequencing of immunological repertoires offers an opportunity for the discovery of novel biomarkers for autoimmune disease. Available bioinformatics techniques however, are not adequately suited for elucidating possible biomarker candidates from within large immunosequencing datasets due to unsatisfactory scalability and sensitivity. Here, we present the Haystack Heuristic, an algorithm customized to computationally extract disease-associated motifs from next-generation-sequenced repertoires by contrasting disease and healthy subjects. This technique employs a local-search graph-theory approach to discover novel motifs in patient data. We apply the Haystack Heuristic to nine million B-cell receptor sequences obtained from nearly 100 individuals in order to elucidate a new motif that is significantly associated with multiple sclerosis. Our results demonstrate the effectiveness of the Haystack Heuristic in computing possible biomarker candidates from high throughput sequencing data and could be generalized to other datasets.