A multi-omics machine learning framework in predicting the recurrence and metastasis of patients with pancreatic adenocarcinoma

Abstract

Local recurrence and distant metastasis are the main causes of death in patients with pancreatic adenocarcinoma (PDAC). Microbial content in PDAC metastasis is still not well-characterized. Here, the tissue microbiome was comprehensively compared between metastatic and non-metastatic PDAC patients. We found that the pancreatic tissue microbiome of metastatic patients was significantly different from that of non-metastatic patients. Further, 10 potential bacterial biomarkers (Kurthia, Gulbenkiania, Acetobacterium and Planctomyces etc.) were identified by differential analysis. Meanwhile, significant differences in expression patterns across multiple omics (lncRNA, miRNA, and mRNA) of PDAC patients were found. The highest accuracy was achieved when these 10 bacterial biomarkers were used as features to predict recurrence or metastasis in PDAC patients, with an AUC of 0.815. Finally, the recurrence and metastasis in PDAC patients were associated with reduced survival and this association was potentially driven by the 10 biomarkers we identified. Our studies highlight the association between the tissue microbiome and recurrence or metastasis of pancreatic adenocarcioma patients, as well as the survival of patients.

Keywords: distant metastasis; local recurrence; microbial community; multi-omics; pancreatic adenocarcinoma; random forest.

FIGURE 1

The correlations of Tumor Node Metastasis classification (TNM) stage with recurrence and metastasis. (A) Patients with recurrence or metastasis are accompanied by increased lymph node involvement. (B) Comparisons of M staging in patients with RM and no-RM; Wilcoxon test is used to compare between different groups of samples. The X-axis represents the different stages of patients; Y-axis represents the recurrence and metastasis, 0: recurrence or metastasis; 1: without recurrence and metastasis.

FIGURE 2

Difference in microbial composition and diversity between the two groups. (A) The top 10 genus levels in two groups; the stacked bar chart showed the composition of patient genus level in two groups of recurrence or metastasis. (B) Comparison of alpha-diversity of two groups based on different indexes. (C) Comparison of beta-diversity of two groups with PCoA. Wilcoxon test was used to detect variation between different groups based on the microbial composition at the genus level. Richness, Chao, and Ace index represent the richness of the microbial species; Shannon, Simpson, and Pielou index represent the diversity of the microbial species; RM, recurrence and metastasis; no-RM, no-recurrence and metastasis.

FIGURE 3

Ten potential biomarkers capable of distinguishing between RM and no-RM. Wilcoxon test was used to detect variation between different groups based on the relative abundance of tissue microbes, When the p-value was less than 0.01, 10 potential genus level microbial markers were identified; the boxplot was used to show the differences between the two groups; RM, recurrence or metastasis; no-RM, without recurrence and metastasis.

FIGURE 4

Difference analysis and enrichment analysis of different omics between two groups. The heat map of the DEGs of (A) lncRNA; (B) miRNA; (C) mRNA between the RM and no-RM group, the x-axis is the sample of two groups, and the y-axis is the top 40 expressions with significant differences screened by DEseq2. (D) GO analysis of DEGs between RM and no-RM. (E) KEGG analysis of the DEGs between RM and no-RM, the X-axis is the ratio of differentially expressed genes enriched in the corresponding pathway, and the Y-axis is the name of the pathway; BP, biological process category; CC, cellular component category; MF, molecular function category; RM, recurrence or metastasis; no-RM, without recurrence and metastasis.

FIGURE 5

Ten identified bacterial biomarkers perform best in predicting recurrence and metastasis in patients with PDAC. (A) Comparison of AUC in patients with recurrence and metastasis predicted by different omics. (B) Evaluation of predictive ability of different evaluation indices for recurrence and metastasis of PDAC patients; micro, microbiome; sig bacteria, 10 identified bacterial biomarkers; sig genes, identified DEGs from mRNA data; AUC, area under curve; ACC, accuracy.

FIGURE 6

Kaplan–Meier survival curve showed significantly different overall survival between RM and no-RM. (A) Relationship between true recurrence and metastasis labels and overall survival of patients. (B) Relationship between recurrence and metastasis labels predicted by the model and the overall survival of patients; RM, recurrence or metastasis; no-RM, without recurrence and metastasis.