Using artificial intelligence for error correction in single cell analyses

The goal of Human Cell Atlas is to remove all the tissues of the human body at different time points, with the goal of creating a reference database for the development of personalized medicine, i.e. to separate healthy from diseased cells.

 This is made possible by the technique known as single-cell RNA sequencing, which helps the researchers to understand that in these small components of life, the gene at any time is stopped or stopped.

 "From the point of view of the method, it represents a huge leap forward.

Read more: Samsung M Series Phones Galaxy M20, Galaxy M10 Launched: Price in India, Specifications

 Earlier, such data could only be obtained from large groups of cells because the measurements require a lot of RNA," Maran Beutner explains.

 "The result was always an average of all used cells, now we are able to get accurate data for each cell," says the doctoral student at Helmholtz Zentrum UNCEN's Institute of Computational Biology (ICB).

Increased sensitivity of technology, however, also means increased sensitivity to the batch effect.

 "Batch describes the fluctuation between the measurement effect, which may be, for example, if the temperature of the device is slightly reduced or the processing time of the cells changes," Maran Butten explains. 

Although many models exist to improve these deviations, those methods are highly dependent on the actual magnitude of the effect. 

"Therefore, we have developed a user-friendly, strong and sensitive measurement called kBET, which determines the amount of difference between experiments and therefore facilitates comparisons of various correction results," says Büttner.

In addition to the batch effect, an event known as a dropout event is a major challenge in single-cell indexing. 

Read more: Computer chip vulnerabilities discovered

"We say that we do a sequence of cells and inspect that a particular genes in the cell do not emit any sign," Dr. Professor of Biological Systems of Mathematical Modeling in Fabian Thes, ICB Director and TUM explains "The root cause of this can be biological or technical in nature: either the gene is not being read by the sequencer because it has not been expressed, or it is not known for technical reasons," he explains.

In order to identify these cases, the biosynthesis of the Theoso group, Gauchen Eraslan and Lucas Simon, used and developed a large number of sequences of several single cells, which is known as an intensive learning algorithm, i.e. artificial intelligence Imitates learning processes in humans (neural network). *

Attracting to a new potential model and comparing the original and discovered data, the algorithm determines whether the absence of gene signals is due to biological or technical failure.

"This model allows even cell type-specific improvements to be artificially similar without two separate cell types," says Fabian Theis.

 "As one of the first deep learning methods in the field of single-cell genomics, the added advantage of the algorithm is that it is well-scalable to handle data sets with millions of cells."

But there is one thing, not the method? Fabian Thes explains, "It is important:" We are not developing software to smooth out the results. Our main goal is to identify and repair errors. 

"We are capable of sharing these figures, which are as accurate as possible with our partners around the world and compare our results with them," - for example when Helmholtz researchers contribute to their algorithms and human cells Analyzes for the atlas because the reliability and data comparison is paramount.