DIC: Deep Imputing and Clustering Single Cell RNA Sequencing Data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) data is prone to missing values due to technical variability.
• Missing values pose a significant challenge for cell type identification and clustering in scRNA-seq analysis.
• Existing imputation-then-clustering methods often fail to leverage cluster structures effectively during imputation.

Purpose of the Study:

• To develop a novel method for simultaneous imputation and clustering of scRNA-seq data.
• To address the limitations of existing methods in exploiting biological cluster structures.
• To improve the accuracy of cell type identification in the presence of missing data.

Main Methods:

• Proposed DIC, a deep neural network with a Y-structure for collaborative imputation and clustering.
• DIC comprises a base module (encoder), an imputation module (decoder), and a clustering module (extra branch).
• The method jointly optimizes imputation and clustering by using cluster structure information for imputation and vice versa.

Main Results:

• Experimental results demonstrate DIC's effectiveness in imputing missing values in scRNA-seq data.
• DIC successfully improves cell type identification through accurate clustering.
• The collaborative approach enhances both imputation accuracy and clustering performance.

Conclusions:

• DIC offers an effective deep learning framework for handling missing data in scRNA-seq analysis.
• The Y-structured network enables simultaneous imputation and clustering, outperforming existing methods.
• DIC provides a robust solution for accurate cell type identification from noisy scRNA-seq datasets.