DeepCpG: accurate prediction of single-cell DNA methylation states using deep learning

Abstract Ependymoma represents a heterogeneous disease affecting the entire neuraxis. Extensive molecular profiling efforts have identified molecular ependymoma subgroups based on DNA methylation. However, the intratumoral heterogeneity and developmental origins of these groups are only partially understood, and effective treatments are still lacking for about 50% of patients with high-risk tumors. We interrogated the cellular architecture of ependymoma using single cell/nucleus RNA-sequencing to analyze 24 tumor specimens across major molecular subgroups and anatomic locations. We additionally analyzed ten patient-derived ependymoma cell models and two patient-derived xenografts (PDXs). Interestingly, we identified an analogous cellular hierarchy across all ependymoma groups, originating from undifferentiated neural stem cell-like populations towards different degrees of impaired differentiation states comprising neuronal precursor-like, astro-glial-like, and ependymal-like tumor cells. While prognostically favorable ependymoma groups predominantly harbored differentiated cell populations, aggressive groups were enriched for undifferentiated subpopulations. Projection of transcriptomic signatures onto an independent bulk RNA-seq cohort stratified patient survival even within known molecular groups, thus refining the prognostic power of DNA methylation-based profiling. Furthermore, we identified novel potentially druggable targets including IGF- and FGF-signaling within poorly prognostic transcriptional programs. Ependymoma-derived cell models/PDXs widely recapitulated the transcriptional programs identified within fresh tumors and are leveraged to validate identified target genes in functional follow-up analyses. Taken together, our analyses reveal a developmental hierarchy and transcriptomic context underlying the biologically and clinically distinct behavior of ependymoma groups. The newly characterized cellular states and underlying regulatory networks could serve as basis for future therapeutic target identification and reveal biomarkers for clinical trials.

Download Full-text

Investigation and Highly Accurate Prediction of Missed Tryptic Cleavages by Deep Learning

Journal of Proteome Research ◽

10.1021/acs.jproteome.1c00346 ◽

2021 ◽

Author(s):

Bo Sun ◽

Pawel Smialowski ◽

Tobias Straub ◽

Axel Imhof

Keyword(s):

Deep Learning ◽

Accurate Prediction

Download Full-text

Fast and accurate prediction of failure pressure of oil and gas defective pipelines using the deep learning model

Reliability Engineering & System Safety ◽

10.1016/j.ress.2021.108016 ◽

2021 ◽

pp. 108016

Author(s):

Yue Su ◽

Jingfa Li ◽

Bo Yu ◽

Yanlin Zhao ◽

Jun Yao

Keyword(s):

Deep Learning ◽

Oil And Gas ◽

Learning Model ◽

Accurate Prediction ◽

Failure Pressure ◽

Deep Learning Model

Download Full-text

Deep Learning Powered Single Cell Biological Microlasers

10.1364/cleo_si.2021.sw3d.4 ◽

2021 ◽

Author(s):

Zhen Qiao ◽

Wen Sun ◽

Randall Ang Jie ◽

Yu-Cheng Chen

Keyword(s):

Deep Learning ◽

Single Cell

Download Full-text

SMILE: Mutual Information Learning for Integration of Single-cell Omics Data

Bioinformatics ◽

10.1093/bioinformatics/btab706 ◽

2021 ◽

Author(s):

Yang Xu ◽

Priyojit Das ◽

Rachel Patton McCord

Keyword(s):

Deep Learning ◽

Mutual Information ◽

Single Cell ◽

Learning Algorithm ◽

Cellular Systems ◽

Supplementary Information ◽

Omics Data ◽

Learning Approaches ◽

Rna Seq ◽

Integrate Data

Abstract Motivation Deep learning approaches have empowered single-cell omics data analysis in many ways and generated new insights from complex cellular systems. As there is an increasing need for single cell omics data to be integrated across sources, types, and features of data, the challenges of integrating single-cell omics data are rising. Here, we present an unsupervised deep learning algorithm that learns discriminative representations for single-cell data via maximizing mutual information, SMILE (Single-cell Mutual Information Learning). Results Using a unique cell-pairing design, SMILE successfully integrates multi-source single-cell transcriptome data, removing batch effects and projecting similar cell types, even from different tissues, into the shared space. SMILE can also integrate data from two or more modalities, such as joint profiling technologies using single-cell ATAC-seq, RNA-seq, DNA methylation, Hi-C, and ChIP data. When paired cells are known, SMILE can integrate data with unmatched feature, such as genes for RNA-seq and genome wide peaks for ATAC-seq. Integrated representations learned from joint profiling technologies can then be used as a framework for comparing independent single source data. Supplementary information Supplementary data are available at Bioinformatics online. The source code of SMILE including analyses of key results in the study can be found at: https://github.com/rpmccordlab/SMILE.

Download Full-text