Cancer type classification using plasma cell-free RNAs derived from human and microbes

Shanwen Chen, Yunfan Jin, Siqi Wang, Shaozhen Xing, Yingchao Wu, Yuhuan Tao, Yongchen Ma, Shuai Zuo, Xiaofan Liu, Yichen Hu, Hongyan Chen, Yuandeng Luo, Feng Xia, Chuan‐Ming Xie, Jianhua Yin, Xin Wang, Zhihua Liu, Ning Zhang, Zhenjiang Zech Xu, Zhi John Lu, Pengyuan WangeLife 2022 · DOI

TL;DR

This study leverages plasma cell-free RNAs (cfRNAs) from both human and microbial sources to develop a more accurate method for detecting and classifying various cancer types. By integrating these dual sources, researchers achieved a significant improvement in cancer detection accuracy and the ability to pinpoint primary tumor locations, offering a promising avenue for non-invasive cancer diagnostics.

What Problem Does This Paper Address?

The early and accurate detection of cancer, along with the identification of its primary site, remains a critical challenge in clinical oncology. While liquid biopsies using cell-free DNA (cfDNA) have shown promise, the utility of cell-free RNA (cfRNA) in plasma for cancer diagnostics is still being explored. Furthermore, the role of microbial cfRNAs in plasma, which are often overlooked or treated as contaminants, has not been fully integrated into cancer detection strategies. This paper aims to bridge this gap by investigating the combined diagnostic power of human and microbial cfRNAs in plasma for detecting and classifying five major cancer types.

Methods

The researchers employed a SMART-based total RNA sequencing (SMART-total) method to profile plasma cfRNAs. This technique is optimized for low-input and potentially degraded RNA samples, making it suitable for plasma cfRNA analysis. A key aspect of their approach was the development of a comprehensive bioinformatic pipeline designed to meticulously filter out potential contaminations from laboratory reagents and the environment. This pipeline involved:

  • Adapter trimming and quality control: Standard preprocessing steps to clean the sequencing data.
  • Mapping to human and microbial genomes: Reads were aligned to the human genome, circular RNA back-spliced junctions, and microbial genomes (bacteria and viruses).
  • K-mer and alignment-based classification: Two distinct methods were used for microbial identification to ensure robustness.
  • Contamination filtering: Rigorous exclusion of reads from known laboratory contaminants, control samples, and common skin microbes.
  • Differential analysis: Using edgeR to identify significantly altered gene and genus abundances between cancer patients and healthy donors, as well as cancer-type specific features.
  • Machine learning for classification: Bootstrapping and balanced random forest classifiers were used to evaluate the performance of human and microbial cfRNA features for binary (cancer vs. healthy) and multiclass (cancer type) classification.

Datasets

The study utilized a cohort of approximately 300 plasma samples, including:

  • Cancer Patients: Samples from 5 major cancer types: colorectal cancer, stomach cancer, liver cancer, lung cancer, and esophageal cancer. The majority of these patients were in early stages (I and II).
  • Healthy Donors (HDs): Control samples from individuals without cancer.

The sequencing data generated in this study has been deposited in the Gene Expression Omnibus (GEO) under accession code GSE174302. A previously published dataset (GSE142987) was also used for validation.

Key Insights and Ideas

  • Microbial cfRNAs are not just noise: The study strongly demonstrates that microbial cfRNAs in plasma are consistently detectable and, crucially, carry clinically relevant information for cancer detection, rather than being solely artifacts of contamination.
  • Dual-source signatures are powerful: Combining human and microbial cfRNA profiles significantly enhances the accuracy of cancer detection and classification compared to using either source alone. This suggests a complex interplay between the host and its microbiome in the context of cancer.
  • Immune suppression signals: Downregulated human gene pathways related to immune function and the increased prevalence of Torque Teno Viruses (TTVs) in cancer patients point towards a potential link between cancer and an immunosuppressed state.
  • Cancer type specificity extends to microbes: Just as human cfRNAs show specificity for different cancer types, so do microbial cfRNAs, offering a novel layer of diagnostic information.

Key Results

  • Cancer vs. Healthy Donor Classification: The combined human and microbial cfRNA profiles achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of approximately 0.9 for distinguishing cancer patients from healthy donors.
  • Cancer Type Classification:
    • Using human cfRNA features alone, the average recall for classifying the five cancer types was 52.5%.
  • Using microbial cfRNA features alone (k-mer based), the average recall was 58.4%.
  • Combining human and microbial features significantly improved the average recall to 60.4%, an 8% increase compared to human features alone.
  • The average top 2 recall also improved from 7.9% (human alone) to 6.2% (combined).
  • Identification of Specific Biomarkers: The study identified specific human genes (e.g., ADAM10, TMEM165) and microbial genera (e.g., Orthohepadnavirus, TTVs, Mycoplasma) associated with particular cancer types.

Limitations and Future Work

  • Cohort Size and Diversity: While the study included ~300 samples, the authors acknowledge that a larger, more carefully designed cohort is necessary for clinical application. They also note potential imbalances in sex distribution for certain cancer types.
  • Computational Mitigation: The researchers employed computational methods (RUVg) to mitigate batch effects and confounding factors, but acknowledge that these are complex issues in high-throughput data.
  • Clinical Validation: The findings provide strong evidence for the clinical relevance and potential utility of cfRNAs, but further validation in larger, prospective clinical trials is essential before widespread clinical adoption.
  • Understanding Microbial Roles: While microbial cfRNAs show diagnostic value, their precise biological roles in cancer development and progression require further investigation. The study highlights the potential for discovering understudied viruses.

Summary updated

Abstract

The utility of cell-free nucleic acids in monitoring cancer has been recognized by both scientists and clinicians. In addition to human transcripts, a fraction of cell-free nucleic acids in human plasma were proven to be derived from microbes and reported to have relevance to cancer. To obtain a better understanding of plasma cell-free RNAs (cfRNAs) in cancer patients, we profiled cfRNAs in ~300 plasma samples of 5 cancer types (colorectal cancer, stomach cancer, liver cancer, lung cancer, and esophageal cancer) and healthy donors (HDs) with RNA-seq. Microbe-derived cfRNAs were consistently detected by different computational methods when potential contaminations were carefully filtered. Clinically relevant signals were identified from human and microbial reads, and enriched Kyoto Encyclopedia of Genes and Genomes pathways of downregulated human genes and higher prevalence torque teno viruses both suggest that a fraction of cancer patients were immunosuppressed. Our data support the diagnostic value of human and microbe-derived plasma cfRNAs for cancer detection, as an area under the ROC curve of approximately 0.9 for distinguishing cancer patients from HDs was achieved. Moreover, human and microbial cfRNAs both have cancer type specificity, and combining two types of features could distinguish tumors of five different primary locations with an average recall of 60.4%. Compared to using human features alone, adding microbial features improved the average recall by approximately 8%. In summary, this work provides evidence for the clinical relevance of human and microbe-derived plasma cfRNAs and their potential utilities in cancer detection as well as the determination of tumor sites.
BibTeX
@inproceedings{chen2022cancer,
  author = {Shanwen Chen and Yunfan Jin and Siqi Wang and Shaozhen Xing and Yingchao Wu and Yuhuan Tao and Yongchen Ma and Shuai Zuo and Xiaofan Liu and Yichen Hu and Hongyan Chen and Yuandeng Luo and Feng Xia and Chuan‐Ming Xie and Jianhua Yin and Xin Wang and Zhihua Liu and Ning Zhang and Zhenjiang Zech Xu and Zhi John Lu and Pengyuan Wang},
  title = {{Cancer type classification using plasma cell-free RNAs derived from human and microbes}},
  booktitle = {eLife},
  year = {2022},
  doi = {10.7554/elife.75181},
}
Figure