Editorial: Artificial intelligence for extracting phenotypic features and disease subtyping applied to single-cell sequencing data

Editorial on the Research Topic
Artificial intelligence for extracting phenotypic features and disease subtyping applied to single-cell sequencing data

With the advent of single-cell sequencing mechanisms, different kinds of omics data (viz., epigenomic, genomic or transcriptomic data profiles) consisting of a lot of individual cells in the parallel-basis, have been generated. Major challenge of the single-cell mechanisms is complexity of genomic or epigenomic information that creates the problem of heterogeneity generating big obstacles to obtain best possible outcomes (or, prediction) in various diseases, mainly for precision oncology. Thus, new artificial intelligence (AI) methods are increasingly proposed and applied to exploit the useful information gathered in those data yielding better accuracy, flexibility, user-friendliness as well as scalability, and finally deliver the effective medicine of precision useful to the patients of the disease. However, the major goal of this Research Topic is to provide the opportunity to the computational biologists and clinical researchers from around the world to contribute their innovative idea and experience towards various underlying problems and their potential solutions in the sphere of computational research in next-generation sequencing data especially, single cell sequencing data for numerous complex diseases (Mallik and Zhao).

Our Research Topic basically covers several new AI tools and frameworks applied to single-cell sequencing data where various computational problems regarding transcriptomics (viz., gene-gene interactions, cell type detection, cell-cell interactions, biomarker discovery, disease classification, disease subtyping and clinical diagnosis) are explored with the advanced translational twist. More emphases are provided to sort out complex real-life medically relevant problems and disorders. Altogether, these manuscripts present kinds of cutting-edge subjects in single-cell sequencing, suggesting that single-cell sequencing data analysis has an increasingly significant role in advancing disease medical application of computational biology. Additionally, it has been observed that new artificial intelligence algorithms are applied to single-cell sequencing data to resolve multiple complex problems that tend to optimize them.

Ligand-Receptor (LR) Hunting methodology employing Random Forest classifier model was developed to discover cell-cell interactions based on single-cell sequencing (scRNA-seq) gene expression data (Lu et al.). LR Hunting discovered the validated interactions between the myeloid cells and CD4⁺ T-cell in Triple Negative Breast Cancer (TNBC). Also, a convolutional neural network (CNN)-based framework for subtype identification was developed where both the transcriptome and methylome profiles were applied to construct the internal classification framework, that clearly separated three subtypes of Glioblastoma (GBM) with higher classification accuracy (Munquad et al.). Another work concerning subtype-specific predictive biomarker discovery was conducted that is applicable to the disease diagnosis and treatment (Munquad et al.). However, it is well-known that identifying potential biomarkers in every cell cluster is inconvenient and unfamiliar that obstructs the systematic study and analysis of single-cell sequencing data. Therefore, to resolve the underlying challenge, a regularized multi-task learning (RMTL) based framework was developed for the simultaneously prediction of the subpopulation related to a specified cell type (Upadhyay and Ray). The regularization strategy was utilized to modulate the multi-task model, smoothing the loss function and thus minimizing the time complexity of the model. Interestingly, artificial neural network (ANN)-based deep learning model was proposed to classify breast cancer samples where the performance of various well-known classifiers or related strategies were evaluated and compared accordingly (Jia et al.). Furthermore, another work concerning cluster-specific frequent biomarker identification through the consecutive utilization of two strategies (viz., dimensionality reduction and Louvain hierarchical agglomerative clustering) was performed from scRNA-seq data (Seth et al.).

Differential network analysis can be able to learn how gene-gene interactions will change along with different biological conditions. A sparse hierarchical Bayesian factor model was developed for scRNA-seq data from different biological conditions, while the methodology utilized a latent factor structure to make impact on the gene expression for cells for helping account for the zero-inflation as well as high cell-to-cell variability (Sekula et al.). Another computational framework was to investigate the molecular interaction techniques in melanoma where a melanoma-specific cell–cell interaction network was built, and a consensus clustering based on network embedding had been applied (Wang et al.).

Finally, it is noticed that majority of the published works in this Research Topic basically cover the drop-out finding, cell-cell interactions, cancer classification model, gene regulatory network inference, cluster-specific frequent biomarker discovery and multimodal data integration. The developed software, tools or methods facilitate the interpretation of the multi-omics data profiles that can improve both the industrial and academic sectors. Moreover, it is highly expected that the upcoming computational tools will be benevolent for biomedical and clinical researchers to understand the applications of AI and machine learning to enhance the quality of research in complex disease detection as well as solve other critical emergency clinical issues (Mallik et al.).

Author contributions

SM and AL wrote and reviewed the editorial of this Research Topic with contributions from all authors.

Acknowledgments

We are grateful to the authors for source of all articles in this Research Topic.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.