Identifikace nových prognostických subtypů chronické lymfocytární leukemie použitím dráhového mutačního skóre a strojového učení
| Title in English | Identification of new prognostic subtypes of chronic lymphocytic leukemia using pathway mutation score and machine learning |
|---|---|
| Authors | |
| Year of publication | 2019 |
| Type | Conference abstract |
| MU Faculty or unit | |
| Citation | |
| Description | Chronic lymphocytic leukemia (CLL) is the most common adult leukemia with a diverse clinical course, largely related to its marked genetic heterogeneity. To date, only a few recurrently mutated genes of clinical relevance have been identified in CLL; Numerous mutations in a number of other genes are usually considered concomitant and their clinical relevance remains unclear. The aim of this study was to describe the effect of these putative accompanying mutations using pathway mutation scores and machine learning methods. Material and methods: We obtained data from 316 CLL patients with somatic hypermutations in immunoglobulin heavy chain (IGHV) from a publicly available database. In this group, we collected 4,739 genes affected by a non-synonymous point mutation and / or a reading frame shift mutation. Subsequently, we determined the biological pathways containing the affected genes and calculated the pathway mutation score for each patient based on the extent of pathway damage. We then used machine learning to identify subgroups of patients with different time to treatment (TTT) and developed a classification model to include patients in these subgroups. Results: We identified five subgroups differing in TTT (p <0.0001). Four of them were characterized by different affected biological processes - cell adhesion, calcium-regulated signaling, synapse organization and ABC protein transport. In the last subgroup, characteristic recurrently affected pathways were not identified. Subsequently, we created a classification model that was able to determine the found subgroups based on a pathway mutation score with an accuracy of 0.77 area under the ROC curve (AUC). We used this model to evaluate our results on an independent dataset of 187 patients with somatic hypermutations in IGHV, in which we identified a similar distribution of subgroups based on our proposed mutation score. Conclusion: In a group of CLL patients with somatic hypermutations IGHV we identified prognostic subgroups based on the pathway mutation score. We believe that these results will help refine the diagnosis of CLL patients. In the future, we are going to apply the proposed procedure to our own set of whole-sequence sequencing data in CLL patients and to investigate the effect of accompanying mutations on clonal evolution of CLL. |
| Related projects: |