Intrinsically Bayesian robust classifier for single-cell gene expression trajectories in gene regulatory networks

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Traditional expression-based phenotype classification lacks specificity due to unrevealed pathway timing and averaged cell expressions.
• Single-cell measurements, when sampled at a sufficient rate, can reveal regulatory timing through expression trajectories.

Purpose of the Study:

• To develop a robust classifier for gene regulatory networks using expression trajectories.
• To discriminate between wild-type and mutated networks, enabling accurate phenotype classification.

Main Methods:

• Modeled network regulation using a Boolean network with perturbation, incorporating biological randomness.
• Developed an intrinsically Bayesian robust classifier for discriminating between networks based on expression trajectories.
• Tested the classifier on a mammalian cell-cycle model, distinguishing normal and mutated gene p27 networks.

Main Results:

• The classifier minimizes expected error across an uncertainty class of Gaussian observation models.
• Simulations demonstrated that reduced perturbation probability and longer trajectories decrease classification error.
• The classifier effectively discriminated between normal and cancerous (p27 mutated) cell-cycle phenotypes.

Conclusions:

• Expression trajectories offer more informative data than average-expression measurements for classification.
• The developed Bayesian robust classifier enhances phenotype classification accuracy in gene regulatory networks.
• Future work should extend methods for obtaining prior distributions to expression trajectories.