Integrating representation learning, permutation, and optimization to detect lineage-related gene expression patterns

Area of Science:

• Computational Biology
• Genomics
• Developmental Biology

Background:

• Recent advances in barcoding technologies enable lineage tree reconstruction alongside paired single-cell RNA sequencing (scRNA-seq).
• These datasets offer a unique opportunity to study gene expression memory maintenance across lineage branching.

Purpose of the Study:

• To develop a computational method for identifying lineage-informative genes and subtrees where lineage and gene expression are tightly coupled.
• To provide a tool for understanding cell state memory maintenance through cell divisions.

Main Methods:

• Developed Permutation, Optimization, and Representation learning based single Cell gene Expression and Lineage ANalysis (PORCELAN).
• Validated PORCELAN using synthetic data.
• Applied PORCELAN to paired lineage and scRNA-seq data from mouse lung cancer, mouse embryogenesis, and C. elegans embryogenesis.

Main Results:

• PORCELAN successfully identified lineage-informative genes and subtrees.
• The method pinpointed subtrees associated with metastasis and new cell state formation in lung cancer.
• Identified genes overlapped with known lung cancer progression pathways.
• Highlighted differences in gene expression memory maintenance between cancer and embryogenesis.

Conclusions:

• PORCELAN is an effective tool for identifying lineage-expression coupled genes and subtrees.
• The findings provide insights into cell state memory during cell division in various biological systems.
• This method can advance the study of cellular dynamics in both development and disease.