scPreGAN, a deep generative model for predicting the response of single-cell expression to perturbation

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables studying cellular responses to perturbations.
• Predicting these responses computationally is crucial due to challenges in experimental cell collection.
• Existing prediction tools have limited accuracy.

Purpose of the Study:

• To develop a novel deep generative model for accurate prediction of single-cell expression responses to perturbations.
• To address the limitations of current in silico prediction methods for cell-level drug responses.

Main Methods:

• Proposed scPreGAN (Single-Cell data Prediction base on GAN), integrating an autoencoder and a generative adversarial network.
• The autoencoder extracts common features from unperturbed and perturbed single-cell data.
• The generative adversarial network predicts the perturbed cell expression profiles.

Main Results:

• scPreGAN demonstrated superior performance compared to three state-of-the-art methods on three real-world datasets.
• The model effectively captures complex gene expression distributions.
• Generated predictions exhibit similar expression abundance to real perturbed data.

Conclusions:

• scPreGAN offers a significant advancement in predicting single-cell responses to perturbations.
• The model's accuracy and ability to generate realistic data hold promise for drug discovery and personalized medicine.
• The scPreGAN implementation is publicly available for research use.