Nature sub-journal: AI algorithm deciphers the genetic characteristics of cancer cells with an accuracy of 99%!

AI has made great achievements again.

This time, a new AI machine learning algorithm "Ikarus" can decipher the differences in genetic characteristics between cancer cells and normal cells.

This research was completed by the team of MDC bioinformatician Altuna Akalin and published in the Nature sub-journal "Genome Biology".

​

Paper address: https://genomebiology.biomedcentral.com/ articles/10.1186/s13059-022-02683-1#Sec8

In addition, the institution responsible for this research, MDC (Max Delbrück center), is also one of the four major research institutions in Germany One of the 16 research centers of the Helmholtz Association.

Since it has such a big background, why is this research so important?

If you select a "common characteristic" from the vast data set, humans will definitely not be able to compare with AI.

To distinguish cancer cells from normal cells, it is necessary to screen out the common characteristics between them.

The Ikarus developed by the MDC research team discovered the common pattern (Pattern) in tumor cells, which consists of a series of genomic features and is common in Various types of cancer.

In addition, the algorithm also detected gene types that had never been linked to cancer.

So the research team asked a simple question:

Is it possible to make a classifier that distinguishes tumor cells from multiple Are normal cells correctly differentiated between cancer types?

So Ikarus was born. It consists of two steps:

1. Discover comprehensive tumor cell characteristics in the form of gene sets by integrating multiple professionally annotated single cell data sets;

2. Train a robust logistic regression classifier to strictly distinguish tumor from normal cells, and then use a customized cell-cell network for network-based propagation of cell labels.

Team leader Altuna Akalin said:

To develop a powerful, sensitive and reproducible computerized tumor cell sorter, we Ikarus has been tested on multiple single-cell data sets of various cancer types obtained using different sequencing technologies to determine its suitability for different experimental settings.

Amazing success rate

Jan Dohmen, first author of the paper, said that after experts have clearly distinguished between healthy cells and In the case of cancer cells, obtaining suitable training data is a major challenge.

Single-cell sequencing data sets are often complex.

This means that the information they contain about the molecular characteristics of individual cells is not very precise, because different numbers of genes are detected in each cell, or because the samples are not processed in the same way. Always the same.

said Dohmen and Dr. Vedran Franke, co-lead of the study,

We screened numerous publications and contacted quite a few research groups to obtain a sufficient data set. The team ultimately chose data from lung cancer and colorectal cancer cells to train the algorithm, which was then applied to data sets from other types of tumors.

During the training phase, Ikarus must find a "signature gene list" and then use it to classify cells.

We tried and refined various methods, and Ikarus ended up using two lists: one for cancer genes and one for genes from other cells genes, Frank explains.

After training, the algorithm can differentiate between healthy and tumor cells in other types of cancer, such as tissue samples from patients with liver cancer or neuroblastoma.

The results in other samples are exciting, and the success rate is surprisingly high, up to 99%.

# "We didn't expect that there would be a common characteristic that could define tumor cells in different types of cancer so precisely," Akalin said.

"But we still can't say whether this approach will work for all types of cancer," Dohmen added.

Not Just Cancer Cell Differentiation

To turn Ikarus into a reliable cancer diagnostic tool, researchers now hope to test it on other types of tumors .

In initial tests, Ikarus has shown that the method can also distinguish other types (and certain subtypes) of cells from tumor cells, and not just tumors Cell detection.

It can be used to detect any cell state, such as cell type, the only requirement is that the cell state exists in at least two independent experiments.

#We want to make this method more comprehensive, develop it further so that it can distinguish all possible cells in a biopsy, Akalin said type.

Application of automated tumor classification on spatial sequencing datasets enables direct annotation of histological samples, thereby facilitating automated digital pathology.

In hospitals, pathologists often just examine tissue samples of tumors under a microscope to identify various cell types. This is a time-consuming and laborious task.

With Ikarus, this step could one day become a fully automated process.

Additionally, Akalin noted, the data can be used to draw conclusions about the tumor's immediate environment. This can help doctors choose the best treatment. The composition of cancer tissue and microenvironment often indicates whether a treatment or drug is effective.

In addition, artificial intelligence may also help develop new drugs.

"Ikarus allows us to identify genes that may contribute to cancer and then target these molecular structures with new therapeutic agents," Akalin said.

The above is the detailed content of Nature sub-journal: AI algorithm deciphers the genetic characteristics of cancer cells with an accuracy of 99%!. For more information, please follow other related articles on the PHP Chinese website!