Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)

Degree Discipline

Electrical Engineering


Single-cell sequencing is a recently advanced revolutionary technology which enables researchers to obtain genomic, transcriptomic, or multi-omics information through gene expression analysis. It gives the advantage of analyzing highly heterogenous cell type information compared to traditional sequencing methods, which is gaining popularity in the biomedical area. Moreover, this analysis can help for early diagnosis and drug development of tumor cells, and cancer cell types. In the workflow of gene expression data profiling, identification of the cell types is an important task, but it faces many challenges like the curse of dimensionality, sparsity, batch effect, and overfitting. However, these challenges can be overcome by performing a feature selection technique which selects more relevant features by reducing feature dimensions. In this research work, recurrent neural network-based feature selection model is proposed to extract relevant features from high dimensional, and low sample size data. Moreover, a deep learning-based gene embedding model is also proposed to reduce data sparsity of single-cell data for cell type identification. The proposed frameworks have been implemented with different architectures of recurrent neural networks, and demonstrated via real-world micro-array datasets and single-cell RNA-seq data and observed that the proposed models perform better than other feature selection models. A semi-supervised model is also implemented using the same workflow of gene embedding concept since labeling data is very cumbersome, time consuming, and requires manual effort and expertise in the field. Therefore, different ratios of labeled data are used in the experiment to validate the concept. Experimental results show that the proposed semi-supervised approach represents very encouraging performance even though a limited number of labeled data is used via the gene embedding concept. In addition, graph attention based autoencoder model has also been studied to learn the latent features by incorporating prior knowledge with gene expression data for cell type classification.

Index Terms — Single-Cell Gene Expression Data, Gene Embedding, Semi-Supervised model, Incorporate Prior Knowledge, Gene-gene Interaction Network, Deep Learning, Graph Auto Encoder

Committee Chair/Advisor

Xiangfang Li

Committee Co-Chair:

Xishuang Dong

Committee Member

John Fuller

Committee Member

Lijun Qian

Committee Member

Lin Li


Prairie View A&M University


© 2021 Prairie View A & M University

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Date of Digitization


Contributing Institution

John B Coleman Library

City of Publication

Prairie View





To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.