Protocol to estimate cell type proportions from bulk RNA-seq using DAISM-DNN^XMBD

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Cell type deconvolution from bulk RNA-seq is crucial for understanding cellular heterogeneity in diseases.
• Existing computational protocols can suffer from performance degradation due to batch effects across datasets.
• Accurate cell type proportion estimation is essential for interpreting complex biological samples.

Purpose of the Study:

• To present a novel computational protocol, DAISM-DNN, for robust cell type proportion estimation.
• To address and mitigate the impact of batch effects in bulk RNA-seq data analysis.
• To provide a detailed methodology for training and applying a deep neural network (DNN) model for deconvolution.

Main Methods:

• Preparation of calibrated human blood samples.
• Training a dataset-specific deep neural network (DNN) model.
• Utilizing the trained DNN model for cell type proportion estimation.

Main Results:

• The DAISM-DNN protocol enables robust cell type proportion estimation.
• The method is designed to overcome batch effects inherent in multi-dataset analyses.
• Successful application demonstrated on human blood samples.

Conclusions:

• The DAISM-DNN protocol offers a reliable solution for cell type deconvolution.
• This approach enhances the accuracy of cellular heterogeneity analysis in bulk RNA-seq data.
• The study provides a practical computational tool for researchers in genomics and bioinformatics.