Biophysically Interpretable Inference of Cell Types from Multimodal Sequencing Data.

Area of Science:

• Single-cell genomics
• Computational biology
• Systems biology

Background:

• Multimodal single-cell genomics captures multiple cellular processes simultaneously.
• Current cell clustering methods often treat data modalities independently, ignoring count data properties.
• Understanding cellular heterogeneity requires robust methods for multimodal data analysis.

Approach:

• We developed meK-Means (mechanistic K-Means), a novel algorithm for interpretable cell cluster determination.
• meK-Means integrates multiple data modalities using a unifying model of transcription.
• The approach models shared biophysical states underlying observed gene expression profiles.

Key Points:

• meK-Means effectively clusters cells using unspliced and spliced mRNA count data.
• It leverages causal, physical relationships between RNA processing modalities.
• The method identifies shared transcriptional kinetics driving gene expression heterogeneity.

Conclusions:

• meK-Means offers a principled approach to cell clustering from multimodal data.
• It redefines cell clusters based on the parameters of underlying cellular processes.
• This enables more mechanistic and interpretable studies of cellular heterogeneity.