Explainable machine learning to identify chronic lymphocytic leukemia and medication use based on gut microbiome data

基于肠道微生物组数据的可解释机器学习方法用于识别慢性淋巴细胞白血病和药物使用情况

阅读:2

Abstract

Medication, particularly antibiotics, significantly alters gut microbiome composition, often reducing microbial diversity and affecting host health. Given that the gut microbiome may influence cancer progression, we integrated clinical, shotgun metagenomic, and medication data to assess microbiome composition across diseased and healthy cohorts, as well as the impact of medication on microbiome variation. The study cohorts included patients with chronic lymphocytic leukemia (CLL, n = 85), acute myeloid leukemia (AML, n = 61), myeloid dysplastic syndrome (MDS), and other severe hematological malignancies (n = 104); patients scheduled for elective cardiac surgery (n = 89); and kidney donors (n = 9), all collected as part of a consecutive microbiome sampling effort at Copenhagen University Hospital, Denmark; and healthy individuals (N = 59). First, our analyses revealed similarities in both diversity and composition between microbiomes of patients with CLL and patients prior to elective cardiac surgery, whereas patients with AML and MDS exhibited the least diverse and most distinct microbiomes. Second, when we quantified sources of microbiome variation, the combination of medication, disease, age, and sex accounted for 4% of variation between all cohorts and 10.4% of variation between CLL and pre-cardiac surgery patients only; the two cohorts selected for comparison due to their similar microbiomes. Notably, this left 90%-95% of the variation unexplained, emphasizing the need for better identification of the parts of the microbiome variation impacting health and disease. Third, using a machine learning approach, we validated and further refined the CLL-associated microbiome pattern from our previous studies. Overall, our data provide a foundation for further investigation into disease-specific microbial signatures and the potential interactions between medication, underlying disease, and the microbiome, with the ultimate goal to improve our understanding and clinical management of CLL.IMPORTANCEThis study reveals how disease and medication influence the gut microbiome in patients with chronic lymphocytic leukemia (CLL) when compared to other more severe hematological malignancies, a cohort of patients scheduled for elective cardiac surgery representing a severely diseased nonhematological cohort, and a cohort of healthy individuals. We found that patients with CLL and those scheduled for cardiac surgery had the most similar microbiome diversity and composition. Similarities across very different disease contexts suggest that disease status alone has limited impact. Consistently, across all cohorts, medication, disease, age, and sex together explained only less of microbiome variation, leaving 90%-95% unexplained. This underscores the important need for better identification of factors shaping the microbiome. In addition, we validated a previously published, machine learning-based CLL-associated microbiome signature, demonstrating the robustness of our previous findings differentiating the microbiome signature for CLL as compared to healthy individuals. The findings expand knowledge on how disease states and medical treatments shape gut microbiome composition and diversity, potentially leading to new ways of managing CLL and improving patient outcomes through microbiome signatures.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。