Scalable optimal Bayesian classification of single-cell trajectories under regulatory model uncertainty

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Single-cell gene expression data offers insights into complex diseases like cancer.
• Gene regulatory network (GRN) inference from this data faces significant uncertainty due to cellular complexity.

Purpose of the Study:

• To develop an optimal classification method for single-cell trajectories that accounts for model uncertainty.
• To address the computational limitations of existing methods for large-scale GRNs.

Main Methods:

• Utilized partially-observed Boolean dynamical systems (POBDS) to model GRNs from noisy gene-expression data.
• Derived an exact optimal Bayesian classifier (OBC) for binary classification of single-cell trajectories.
• Introduced a scalable particle-based classification method to overcome OBC's computational and memory constraints.

Main Results:

• Developed a particle-based method that is highly scalable for large GRNs.
• Achieved much lower complexity compared to the optimal Bayesian classifier.
• Demonstrated the method's performance on a T-cell large granular lymphocyte (T-LGL) leukemia GRN model.

Conclusions:

• The proposed particle-based method provides a practical and scalable solution for classifying single-cell trajectories in GRN analysis.
• The method effectively handles uncertainty in gene expression data.
• Numerical experiments validated the performance on a relevant biological network.