A deep generative model for deciphering cellular dynamics and in silico drug discovery in complex diseases

Area of Science:

• Computational biology
• Genomics
• Drug discovery

Background:

• Single-cell transcriptomics offers insights into human diseases but lacks advanced computational tools for disease progression analysis and in silico drug intervention.
• Analyzing complex cellular dynamics in diseases requires sophisticated computational approaches.

Purpose of the Study:

• To introduce UNAGI, a deep generative neural network designed for analyzing time-series single-cell transcriptomic data.
• To enhance disease progression modeling and in silico drug screening capabilities.
• To identify potential therapeutic targets and drug candidates for complex diseases.

Main Methods:

• Development of UNAGI, a deep generative neural network for time-series single-cell transcriptomic data analysis.
• Application of UNAGI to idiopathic pulmonary fibrosis (IPF) patient data to learn disease-informed cell embeddings.
• Validation of UNAGI's predictions using proteomics and human precision-cut lung slice models.
• Testing UNAGI's adaptability on COVID-19 data.

Main Results:

• UNAGI effectively captures complex cellular dynamics in disease progression.
• The tool identified potential therapeutic drug candidates for IPF, including nifedipine with confirmed anti-fibrotic effects in human lung tissues.
• Proteomic validation confirmed the accuracy of UNAGI's cellular dynamics analysis.
• UNAGI demonstrated versatility and applicability to other diseases like COVID-19.

Conclusions:

• UNAGI provides a powerful computational tool for decoding complex cellular dynamics in disease progression.
• It enhances in silico drug perturbation modeling and screening, accelerating therapeutic discovery.
• UNAGI shows broad applicability across diverse pathological landscapes, aiding the search for novel treatments.