16. John Lian, BS

Introduction/Background: Somatic tumor cell mutation and host immune response are predictors of immunotherapy response in cancer patients. These are observed as tumor mutational burden (TMB), obtained from genetic sequencing assays, and tumor-infiltrating lymphocyte (TIL) response, obtained from pathology reports. Associations between TMB and TIL elucidate the role tumor cell mutations play in immune evasion. The narrative, free-text nature of pathology and genetic assay reports requires manual chart review; as such, large patient cohort analyses are lacking. Natural language processing (NLP) enables automatic feature extraction in large cohorts for the determination of TMB and pathological feature associations.

Methods: 11,072 pathology reports and 932 assay reports of melanoma patients were obtained from the same institution. An NLP algorithm identified a cohort of 242 matched reports belonging to the same tumor. TMB and pathology features including TIL status, invasion depth, cell size, and cell type were extracted using NLP; accuracy and precision were subsequently confirmed with manual evaluation. TMB and pathological feature associations were evaluated using one-way ANOVA, unpaired T-test, and Pearson r tests.

Results: TMB values of the 242 cases range from 0.76 to 206.8 (average 19.67). Brisk TIL status is associated with significantly lower TMB (average=10.39, n=9) compared to absent TIL status (average=27.22, n=40) and non-brisk TILs (average=18.46, n=169, p=0.0371). Invasion depth surprisingly does not correlate with TMB (p=0.2831). Cell size is not significantly associated with TMB (p=0.3338). Cell type, specifically indolent desmoplastic melanoma composed of spindle cells, is significantly associated with low TMB (p=0.0114).

Conclusion: Use of NLP in processing unstructured free-text clinical data allows for large-scale analyses. High TMB is inversely correlated with brisk TIL presence, suggesting that increased mutations in melanoma tumors are associated with immune evasion. These findings indicate the need for future studies in interactions between somatic mutations and immunity to enhance immunotherapeutic treatments.