AI facilitates the identification of tissue substructures thanks to the transcriptome

Deciphering tissue substructures or spatial domains is one of the great challenges for biologists. Thus, different methods have been adopted, but few make effective use of available spatial information giving rise to very discrete tissue substructures. To overcome this obstacle, recent studies have been carried out and have resulted in the creation of a new approach based on a well-structured algorithm.

Indeed, the model uses artificial intelligence technology to integrate spatial location information and spatial blot gene expression profile. Therefore, it was possible to better describe the boundary of the cell space domain and then improve the performance of cellular substructuress.

ZHANG Shihua, an expert in machine learning and computational biology, was the lead author of the study. In addition, the research results were published in the journal NatureCommunications.

STARGATE, the complete and reliable next-generation algorithm

In this algorithm, an autoencoder and a graphical attention mechanism are introduced in the middle hidden layer. Thus, these two tools are able to learn heterogeneous similarities between neighboring points in an adaptive way. According to the researchers, the new model converts the spatial location information into a spatial neighbor network between the spatial spots.

Then, this algorithm feeds the gene expression information and the spatial network into a graphical attention autoencoder to learn a low-dimensional representation of spots. In addition, STARGATE combines features of Visium data and suggests a cell type support module based on pre-clustering of expression information.

Therefore, he can further describe the cell space domain boundary. Finally, the new model can reduce the batch effect between different sections by introducing a spatial network between adjacent sections. In this way, an improvement in the performance of the three-dimensional tissue substructures ensues.

A technology with promising potential

Most of the existing clustering methods do not make efficient use of available spatial information, which gives very discreet tissue substructures. The superiority of STARGATE for deciphering tissue substructures or spatial domains has been validated in various datasets.

It should be noted that it can be used for analyze spatial transcriptome data of different sequencing platforms with various spatial resolutions.

“With the rapid development of spatial omics technology and the continued accumulation of data, this new STAGATE model may facilitate the accurate analysis of large-scale spatial transcriptome data and advance our understanding of tissue substructures. »

ZHANG Shihua, expert in machine learning and computational biology, and lead author of the study


We would like to say thanks to the author of this article for this outstanding content

AI facilitates the identification of tissue substructures thanks to the transcriptome

Check out our social media profiles as well as other related pages