In a recent study, researchers have discovered a new method that use brain networks and supervised machine-learning techniques to efficiently identify cell types following single-cell RNA sequencing (scRNA-seq).
Instead of depending on marker genes that are not available for all cell types, scientists at Carnegie Mellon University adopted the new automated technique to evaluate all scRNA-seq data and select just specific parameters required for differentiating one cell from another. The findings could help researchers characterize new cell subtypes and distinguish between healthy and damaged or diseased cells.
According to authors of the study, published in the online journal Nature Communications, ‘scQuery’ is web server that will enable any researcher to use the new technique. While single-cell sequencing has been popularly used as a tool to identify and differentiate cell types, it allowed scientist to process only batches of cells to get the results that revealed an overall average of the value.
The new technique will be utilized as a part of ‘Human BioMolecular Atlas Program’ by National Institutes of Health, which creates a 3D map of the human body, differentiating tissue at cellular level. Amir Alavi, a computational biologist said that each experiment generates enormous data points which is creating a ‘Big Data’ problem that previous methods of analysis cannot manage.
The researchers created an automated pipeline, with the objective to download all the available RNA sequencing data of mice from biggest repositories and to be able to identify genes and proteins expressed in each cell. After labeling each cell according to type, a computational neural network modelled on the human brain was employed to compare the cells and evaluate the factors that distinguish them individually.
To test the new model, the research team utilized RNA sequencing data from an animal study of a cognitive disease equivalent to Alzheimer’s. As expected, same number of brain cells were found in healthy as well as diseased tissue, however, the diseased tissue consisted more number of immune cells as a response to the disease.
The scQuery web server was then developed using the automated pipeline and techniques which promotes comparative analysis of new scRNA-seq data. After entering a single-cell experiment into this server, the neural networks and matching techniques quickly identify related types of cells and even previous results of the similar cells.