PHILADELPHIA, PA — In a groundbreaking development, scientists at the Wistar Institute in Philadelphia have pioneered a new machine learning tool that can identify microbes associated with cancer. Led by Dr. Noam Auslander, assistant professor in the Ellen and Ronald Caplan Cancer Center’s Molecular & Cellular Oncogenesis Program, the team has used this tool to detect biomarkers for esophageal carcinoma (ESCA) through the analysis of short read RNA-sequencing data.
RNA sequencing (RNAseq), a method used to identify mRNA in a population of cells to determine which genes are expressing, is commonly used to analyze tumor microenvironments. Theoretically, RNAseq data can reveal the expression of microbial genes in cancerous tissue, potentially identifying microbiome shifts contributing to the cancer’s development. However, short RNAseq “reads” present a challenge for classification into diverse microbial genetic origins due to their length.
To overcome this hurdle, Dr. Auslander’s team trained a convolutional neural network, a type of machine learning technology capable of self-learning to accurately assess large quantities of data. Using large publicly available datasets of characterized short-read data, they trained the network to sort short-read RNAseq data by its likely origin: human, viral, or bacterial. This approach aimed to reduce the number of short reads requiring assembly for identification, thereby decreasing the computational load of screening for microbial influences in cancer tissue.
Upon training the model, the team selectively analyzed ESCA tissue for reads of microbial origin and compared them with apparently healthy esophageal tissue. They discovered several instances of microbial expression present in ESCA with significantly less incidence in healthy esophageal tissue.
Notably, they found that nearly half of the microbial genes over-expressed in cancer originated from bacteriophages, viruses that infect bacteria, suggesting that viruses infecting microorganisms within the tumor microenvironment could facilitate ongoing cancerous gene expression.
Moreover, the team identified patient outcome predictors in the data. They found bacterial iron-sulfur proteins impacting human genes involved in ferroptosis—a type of cell death pathway modulated by iron—which predicted poor prognosis in ESCA patients. Microbial genes involved in mitochondrial reprogramming were also discovered to influence ESCA patient prognosis.
This innovative research holds promising implications for the medical community, potentially revolutionizing how we understand and approach cancer treatment. As machine learning continues to advance, its applications in biomedical research are showing immense potential, opening new avenues for groundbreaking discoveries.
For the latest news on everything happening in Chester County and the surrounding area, be sure to follow MyChesCo on Google News and MSN.