A deep learning approach reveals unexplored landscape of viral expression in cancer

VirusPredictor: XGBoost-based software to predict virus-related sequences in human data

MOTIVATION

Discovering disease causative pathogens, particularly viruses without reference genomes, poses a technical challenge as they are often unidentifiable through sequence alignment. Machine learning prediction of patient high-throughput sequences unmappable to human and pathogen genomes may reveal sequences originating from uncharacterized viruses. Currently, there is a lack of software specifically designed for accurately predicting such viral sequences in human data.

RESULTS

We developed a fast XGBoost method and software VirusPredictor leveraging an in-house viral genome database. Our two-step XGBoost models first classify each query sequence into one of three groups: infectious virus, endogenous retrovirus (ERV) or non-ERV human. The prediction accuracies increased as the sequences became longer, i.e. 0.76, 0.93, and 0.98 for 150-350 (Illumina short reads), 850-950 (Sanger sequencing data), and 2000-5000 bp sequences, respectively. Then, sequences predicted to be from infectious viruses are further classified into one of six virus taxonomic subgroups, and the accuracies increased from 0.92 to >0.98 when query sequences increased from 150-350 to >850 bp. The results suggest that Illumina short reads should be de novo assembled into contigs (e.g. ∼1000 bp or longer) before prediction whenever possible. We applied VirusPredictor to multiple real genomic and metagenomic datasets and obtained high accuracies. VirusPredictor, a user-friendly open-source Python software, is useful for predicting the origins of patients' unmappable sequences. This study is the first to classify ERVs in infectious viral sequence prediction. This is also the first study combining virus sub-group predictions.

AVAILABILITY AND IMPLEMENTATION

www.dllab.org/software/VirusPredictor.html.

Widespread Horizontal Gene Transfer Among Animal Viruses

The initial objective of this study was to shed light on the evolution of small DNA tumor viruses by analyzing de novo assemblies of publicly available deep sequencing datasets. The survey generated a searchable database of contig snapshots representing more than 100,000 Sequence Read Archive records. Using modern structure-aware search tools, we iteratively broadened the search to include an increasingly wide range of other virus families. The analysis revealed a surprisingly diverse range of chimeras involving different virus groups. In some instances, genes resembling known DNA-replication modules or known virion protein operons were paired with unrecognizable sequences that structural predictions suggest may represent previously unknown replicases and novel virion architectures. Discrete clades of an emerging group called adintoviruses were discovered in datasets representing humans and other primates. As a proof of concept, we show that the contig database is also useful for discovering RNA viruses and candidate archaeal phages. The ancillary searches revealed additional examples of chimerization between different virus groups. The observations support a gene-centric taxonomic framework that should be useful for future virus-hunting efforts.

Single cell deciphering of progression trajectories of the tumor ecosystem in head and neck cancer

Head and neck squamous cell carcinoma is the sixth most common cancer worldwide and has high heterogeneity and unsatisfactory outcomes. To better characterize the tumor progression trajectory, we perform single-cell RNA sequencing of normal tissue, precancerous tissue, early-stage, advanced-stage cancer tissue, lymph node, and recurrent tumors tissue samples. We identify the transcriptional development trajectory of malignant epithelial cells and a tumorigenic epithelial subcluster regulated by TFDP1. Furthermore, we find that the infiltration of POSTN+ fibroblasts and SPP1+ macrophages gradually increases with tumor progression; their interaction or interaction with malignant cells also gradually increase to shape the desmoplastic microenvironment and reprogram malignant cells to promote tumor progression. Additionally, we demonstrate that during lymph node metastasis, exhausted CD8+ T cells with high CXCL13 expression strongly interact with tumor cells to acquire more aggressive phenotypes of extranodal expansion. Finally, we delineate the distinct features of malignant epithelial cells in primary and recurrent tumors, providing a theoretical foundation for the precise selection of targeted therapy for tumors at different stages. In summary, the current study offers a comprehensive landscape and deep insight into epithelial and microenvironmental reprogramming throughout initiation, progression, lymph node metastasis and recurrence of head and neck squamous cell carcinoma.

The Microbiome Matters: Its Impact on Cancer Development and Therapeutic Responses

In the evolving landscape of cancer research, the human microbiome emerges as a pivotal determinant reshaping our understanding of tumorigenesis and therapeutic responses. Advanced sequencing technologies have uncovered a vibrant microbial community not confined to the gut but thriving within tumor tissues. Comprising bacteria, viruses, and fungi, this diverse microbiota displays distinct signatures across various cancers, with most research primarily focusing on bacteria. The correlations between specific microbial taxa within different cancer types underscore their pivotal roles in driving tumorigenesis and influencing therapeutic responses, particularly in chemotherapy and immunotherapy. This review amalgamates recent discoveries, emphasizing the translocation of the oral microbiome to the gut as a potential marker for microbiome dysbiosis across diverse cancer types and delves into potential mechanisms contributing to cancer promotion. Furthermore, it highlights the adverse effects of the microbiome on cancer development while exploring its potential in fortifying strategies for cancer prevention and treatment.

The Breast Microbiome in Breast Cancer Risk and Progression: A Narrative Review

A decade ago, studies in human populations first revealed the existence of a unique microbial community in the breast, a tissue historically viewed as sterile, with microbial origins seeded through the nipple and/or translocation from other body sites. Since then, research efforts have been made to characterize the microbiome in healthy and cancerous breast tissues. The purpose of this review is to summarize the current evidence for the association of the breast microbiome with breast cancer risk and progression. Briefly, while many studies have examined the breast microbiome in patients with breast cancer, and compared it with the microbiome of benign breast disease tissue or normal breast tissue, these studies have varied widely in their sample sizes, methods, and quality of evidence. Thus, while several large and rigorous cross-sectional studies have provided key evidence of an altered microbiome in breast tumors compared with normal adjacent and healthy control tissue, there are few consistent patterns of perturbed microbial taxa. In addition, only one large prospective study has provided evidence of a relationship between the breast tumor microbiota and cancer prognosis. Future research studies featuring large, well-characterized cohorts with prospective follow-up for breast cancer incidence, progression, and response to treatment are warranted.

Leveraging a KRAS-based signature to predict the prognosis and drug sensitivity of colon cancer and identifying SPINK4 as a new biomarker

KRAS is one of the leading mutations reported in colon cancer. However, there are few studies on the application of KRAS related signature in predicting prognosis and drug sensitivity of colon cancer patient. We identified KRAS related differentially expressed genes (DEGs) using The Cancer Genome Atlas (TCGA) database. A signature closely related to overall survival was recognized with Kaplan-Meier survival analysis and univariate cox regression analysis. Then we validated this signature with overall expression score (OE score) algorithm using both scRNA-seq and bulk RNA-seq data. Based on this signature, we performed LASSO cox regression to establish a prognostic model, and corresponding scores were calculated. Differences in genomic alteration, immune microenvironment, drug sensitivity between high- and low-KRD score groups were investigated. A KRAS related signature composed of 80 DEGs in colon cancer were recognized, among which 19 genes were selected to construct a prognostic model. This KRAS related signature was significantly correlated with worse prognosis. Furthermore, patients who scored lower in the prognostic model presented a higher likelihood of responding to chemotherapy, targeted therapy and immunotherapy. Furthermore, among the 19 selected genes in the model, SPINK4 was identified as an independent prognostic biomarker. Further validation in vitro indicated the knockdown of SPINK4 promoted the proliferation and migration of SW48 cells. In conclusion, a novel KRAS related signature was identified and validated based on clinical and genomic information from TCGA and GEO databases. The signature was proved to regulate genomic alteration, immune microenvironment and drug sensitivity in colon cancer, and thus might serve as a predictor for individual prognosis and treatment.

Microbial gene expression analysis of healthy and cancerous esophagus uncovers bacterial biomarkers of clinical outcomes

Local microbiome shifts are implicated in the development and progression of gastrointestinal cancers, and in particular, esophageal carcinoma (ESCA), which is among the most aggressive malignancies. Short-read RNA sequencing (RNAseq) is currently the leading technology to study gene expression changes in cancer. However, using RNAseq to study microbial gene expression is challenging. Here, we establish a new tool to efficiently detect viral and bacterial expression in human tissues through RNAseq. This approach employs a neural network to predict reads of likely microbial origin, which are targeted for assembly into longer contigs, improving identification of microbial species and genes. This approach is applied to perform a systematic comparison of bacterial expression in ESCA and healthy esophagi. We uncover bacterial genera that are over or underabundant in ESCA vs healthy esophagi both before and after correction for possible covariates, including patient metadata. However, we find that bacterial taxonomies are not significantly associated with clinical outcomes. Strikingly, in contrast, dozens of microbial proteins were significantly associated with poor patient outcomes and in particular, proteins that perform mitochondrial functions and iron-sulfur coordination. We further demonstrate associations between these microbial proteins and dysregulated host pathways in ESCA patients. Overall, these results suggest possible influences of bacteria on the development of ESCA and uncover new prognostic biomarkers based on microbial genes. In addition, this study provides a framework for the analysis of other human malignancies whose development may be driven by pathogens.

Profiling of Microbial Landscape in Lung of Chronic Obstructive Pulmonary Disease Patients Using RNA Sequencing

Purpose

The aim of the study was to use RNA sequencing (RNA-seq) data of lung from chronic obstructive pulmonary disease (COPD) patients to identify the bacteria that are most commonly detected. Additionally, the study sought to investigate the differences in these infections between normal lung tissues and those affected by COPD.

Patients and Methods

We re-analyzed RNA-seq data of lung from 99 COPD patients and 93 non-COPD smokers to determine the extent to which the metagenomes differed between the two groups and to assess the reliability of the metagenomes. We used unmapped reads in the RNA-seq data that were not aligned to the human reference genome to identify more common infections in COPD patients.

Results

We identified 18 bacteria that exhibited significant differences between the COPD and non-COPD smoker groups. Among these, Yersinia enterocolitica was found to be more than 30% more abundant in COPD. Additionally, we observed difference in detection rate based on smoking history. To ensure the accuracy of our findings and distinguish them from false positives, we double-check the metagenomic profile using Basic Local Alignment Search Tool (BLAST). We were able to identify and remove specific species that might have been misclassified as other species in Kraken2 but were actually Staphylococcus aureus, as identified by BLAST analysis.

Conclusion

This study highlighted the method of using unmapped reads, which were not typically used in sequencing data, to identify microorganisms present in patients with lung diseases such as COPD. This method expanded our understanding of the microbial landscape in COPD and provided insights into the potential role of microorganisms in disease development and progression.

Overexpression of the mitochondrial anti-viral signaling protein, MAVS, in cancers is associated with cell survival and inflammation

Mitochondrial anti-viral signaling protein (MAVS) plays an important role in host defense against viral infection via coordinating the activation of NF-κB and interferon regulatory factors. The mitochondrial-bound form of MAVS is essential for its anti-viral innate immunity. Recently, tumor cells were proposed to mimic a viral infection by activating RNA-sensing pattern recognition receptors. Here, we demonstrate that MAVS is overexpressed in a panel of viral non-infected cancer cell lines and patient-derived tumors, including lung, liver, bladder, and cervical cancers, and we studied its role in cancer. Silencing MAVS expression reduced cell proliferation and the expression and nuclear translocation of proteins associated with transcriptional regulation, inflammation, and immunity. MAVS depletion reduced expression of the inflammasome components and inhibited its activation/assembly. Moreover, MAVS directly interacts with the mitochondrial protein VDAC1, decreasing its conductance, and we identified the VDAC1 binding site in MAVS. Our findings suggest that MAVS depletion, by reducing cancer cell proliferation and inflammation, represents a new target for cancer therapy.

Gastric Cancer: The Microbiome Beyond Helicobacter pylori

Gastric cancer remains an important global health burden. Helicobacter pylori is the major etiological factor in gastric cancer, infecting the stomach of almost half of the population worldwide. Recent progress in microbiome research offered a new perspective on the complexity of the microbial communities of the stomach. Still, the role of the microbiome of the stomach beyond H. pylori in gastric carcinogenesis is not well understood and requires deeper investigation. The gastric bacterial communities of gastric cancer patients are distinct from those of patients without cancer, but the microbial alterations that occur along the process of gastric carcinogenesis, and the mechanisms through which microorganisms influence cancer progression still need to be clarified. Except for Epstein-Barr virus, the potential significance of the virome and of the mycobiome in gastric cancer have received less attention. This chapter updates the current knowledge regarding the gastric microbiome, including bacteria, viruses, and fungi, within the context of H. pylori-mediated carcinogenesis. It also reviews the possible roles of the local gastric microbiota, as well as the microbial communities of the oral and gut ecosystems, as biomarkers for gastric cancer detection. Finally, it discusses future perspectives and acknowledges limitations in the area of microbiome research in the gastric cancer setting, to which further research efforts should be directed. These will be fundamental not only to increase our current understanding of host-microbial interactions but also to facilitate translation of the findings into innovative preventive, diagnostic, and therapeutic strategies to decrease the global burden of gastric cancer.