New method for analyzing genes activity helps predict cancer patients survival

An international research team developed a new method for determining cell types in a tissue sample. The scientists determined the link between the activity of genes in the same cell type and made a model capable of “recognizing” different cell types in mixed samples based on this relation. This approach works for all tissues, so it can be used to understand, for example, the proportions of which cell types are associated with the survival of patients with different types of cancer. The results are published in Nature Communications.

Image credit: mohamed hassan via Pxhere, CC0 Public Domain

The analysis of a transcriptome, or the set of all RNA molecules in set of samples, is widely used in biomedical research. Using this method, researchers may analyze molecular processes in a tissue and, for instance, characterize the severity of cancer. However, tissue samples can contain millions of different cells that need to be distinguished from each other in order to understand what is happening.

Working on this problem, scientists have developed special deconvolution algorithms. They provide an opportunity to decompose the data and match it to different cell types. It helps to understand what cell types are present in the sample, what is their proportion and how it affects the transcript. However, once the sample contains many different cell types, it becomes difficult to “identify” all of them without any additional information.

The international research group from ITMO University and the University of Washington in St. Louis, US, found a way to overcome this obstacle and proposed the new method for analyzing transcriptome samples. It can determine with high accuracy which cell types the samples contain based on the genes mutual linearity principle: the expression levels of two genes specific for the same cell type linearly depend on each other. The scientists used this relation to construct networks of such linearly dependent genes. By analyzing such networks one can determine what cells are in the samples.

Scientists have shown that all deconvolution algorithms are subject to the same bias: if different cell types in the sample have different RNA amount, all deconvolution algorithms cannot estimate cell type proportions accurately. To test this experimentally, 2 types of cells with different amounts of RNA were selected and mixed in different predetermined proportions. After that researchers used various deconvolution algorithms to determine cell types ratios.

“We saw that existing algorithms would always be wrong about number of cells since they estimate the amount of RNA in the samples. Yet, if measurements are made with adding particular amount of artificial RNA to each sample, the predicted cell types proportions  can be corrected to become more accurate,” explains Konstantin Zaitsev, researcher at the Laboratory of Computer Technologies of ITMO University.

“Our approach is best suited for analyzing mixed samples without sufficient information about their composition. As long the method doesn’t require any “hints”, it can be used for any tissue type. For example, it can detect differences in cell composition after vaccination in blood samples. Moreover, using the TCGA public database (The Cancer Genome Atlas), we are already trying to identify cell types associated with survival of patients with different cancers,” Konstantin concludes.

Reference:

Complete deconvolution of cellular mixtures based on linearity of transcriptional signatures

Konstantin Zaitsev et al.

Nature Communications. 17 May 2019

https://www.nature.com/articles/s41467-019-09990-5

Source: ITMO University