Dr. Dvir Aran from the Faculty of Biology at the Technion has developed a computational method that extracts important information from tilings. RNA At the single cell level
Credit: Technion spokespeople
Characterization of single cells is an important technological challenge in the worlds of biology and medicine, especially in the context of diagnosing diseases. Such characterization allows the identification, at an early stage of the disease, of cellular disruptions indicative of it. One way to characterize a single cell is RNA sequencing (scRNA-seq). Although such sequencing provides a high-resolution "image" of the single cell, the information obtained from it is complex and noisy, making it difficult to refine it for the purpose of diagnosing disruptions and diseases.
A pioneering platform developed by Dr. Dvir Aran from the Faculty of Biology at the Technion provides high-quality integration and analysis of data obtained from multiple single-cell RNA sequencing. The platform, called CellMentor, is presented in the journal Nature Communications.The study was co-authored by Dr. Aran and students Or Havdali, who completed her master's degree in Aran's lab, and Kate Petrenko, who completed her master's degree under the supervision of Dr. Aran and continued her doctorate at the Rappaport Faculty of Medicine.
The platform the three developed, CellMentor, is a machine learning method based on a new approach the researchers call “aware dimensionality reduction.” The tool performs excellently in several aspects:
1. Successful analysis both in complex simulations and on real data.
2. Success in analyzing the sequencing of various cells such as blood cells, pancreas, or melanoma.
3. Good performance even in characterizing cells whose prevalence in a sample is very low (around 1%).
According to Dr. Aran, "This tool is effective not only on a theoretical level. In a separate study in our laboratory, we tested cells from a solid cancer tumor (neuroblastoma) using CellMentor and thus discovered a mechanism that could not be discovered using standard methods. In conclusion, in an article inNature Communications. "We present a new and powerful platform for integrative analyses of scRNA-seq data from diverse sources. This platform is already helping to integrate multiple RNA sequencing, deepen the understanding of complex tissues, and study diseases based on cellular patterns."
The research was supported by the Israel National Science Foundation and the Azrieli Foundation.
To the article in the journal Nature Communications.
More of the topic in Hayadan:
One response
The differences between the models as expected from the graphs are very impressive. It certainly looks like mathematical functionality on noisy biological data. I wonder what dimension is consciously removed and what are the criteria for selecting it. Apparently the changes are dramatic.
It will probably be possible to adapt the idea to other worlds where separations are difficult. Well done, keep up the good work.