Embedding of Genes Using Cancer Gene Expression Data: Biological Relevance and Potential Application on Biomarker Discovery

Abstract

Artificial neural networks (ANNs) have been utilized for classification and prediction task with remarkable accuracy. However, its implications for unsupervised data mining using molecular data is under-explored. We found that embedding can extract biologically relevant information from The Cancer Genome Atlas (TCGA) gene expression dataset by learning a vector representation through gene co-occurrence. Ground truth relationship, such as cancer types of the input sample and semantic meaning of genes, were showed to retain in the resulting entity matrices. We also demonstrated the interpretability and usage of these matrices in shortlisting candidates from a long gene list as in the case of immunotherapy response. 73 related genes are singled out while the relatedness of 55 genes with immune checkpoint proteins (PD-1, PD-L1, and CTLA-4) are supported by literature. 16 novel genes (ACAP1, C11orf45, CD79B, CFP, CLIC2, CMPK2, CXCR2P1, CYTIP, FER, MCTO1, MMP25, RASGEF1B, SLFN12, TBC1D10C, TRAF3IP3, TTC39B) related to immune checkpoint proteins were identified. Thus, this method is feasible to mine big volume of biological data, and embedding would be a valuable tool to discover novel knowledge from omics data. The resulting embedding matrices mined from TCGA gene expression data are interactively explorable online (http://bit.ly/tcga-embedding-cancer) and could serve as an informative reference for gene relatedness in the context of cancer and is readily applicable to biomarker discovery of any molecular targeted therapy.

Tre2-Bub2-Cdc16 Family Proteins Based Nomogram Serve as a Promising Prognosis Predicting Model for Melanoma

“…Knockdown of TBC1D7 resulted in the activation of mTORC1 signaling, and enhanced cell growth (26). Interests in TBC1D10C are mainly focused on immune response (27,28). Two pieces of literature reported TBC1D19, one showed that TBC1D19 acted upon cell polarity and decreased TBC1D19 expression contributed to the disruption of odontoblast polarity and apoptosis (29).…”

Factorized embeddings learns rich and biologically meaningful embedding spaces using factorized tensor decomposition

“…Recently, Schreiber et al have published Avocado, a deep neural network tensor factorization tool specialized in epigenomics data, that learns a representation of the human genome, allowing for imputation of epigenomics data and other related tasks ( Schreiber et al , 2019 ). Lastly, similar work by Choy et al showcased a shallow artificial neural network (ANN) to represent genes and samples in high-dimensional embedding spaces, while simultaneously extracting information about genes and samples ( Choy et al , 2019 ). Choy et al reported that they were able to cluster cancers according to gene expression into meta-groups that may then be used for predicting immune checkpoint therapy responders ( Choy et al , 2019 ).…”

“…Lastly, similar work by Choy et al showcased a shallow artificial neural network (ANN) to represent genes and samples in high-dimensional embedding spaces, while simultaneously extracting information about genes and samples ( Choy et al , 2019 ). Choy et al reported that they were able to cluster cancers according to gene expression into meta-groups that may then be used for predicting immune checkpoint therapy responders ( Choy et al , 2019 ).…”

“…In principle, the model proposed by Choy et al (2019) yields an attractive for data-mining double representation of genes and samples. In this aticle, we extend the idea behind simultaneously training a distributed representation for genes and samples, presented in Choy et al (2019) to the notion of factorized embeddings (FE), an ANN that learns independent embedding spaces to represent factors of RNA-Seq data.…”

“…In principle, the model proposed by Choy et al (2019) yields an attractive for data-mining double representation of genes and samples. In this aticle, we extend the idea behind simultaneously training a distributed representation for genes and samples, presented in Choy et al (2019) to the notion of factorized embeddings (FE), an ANN that learns independent embedding spaces to represent factors of RNA-Seq data. We present the general framework of the FE model and compare it, when possible, to the models from Choy et al (2019) and Du et al (2019) , as well as other standard dimensionality reduction algorithms such as t-SNE, UMAP and PCA.…”