Branching topology of the human embryo transcriptome revealed by Entropy Sort Feature Weighting

Area of Science:

• Computational Biology
• Developmental Biology
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) analysis commonly relies on selecting highly variable genes (HVGs) for feature selection.
• Existing methods may obscure subtle biological signals, particularly in complex developmental processes.
• A robust feature selection framework is needed to accurately capture cell-state dynamics.

Purpose of the Study:

• To introduce and validate Entropy Sorting as a novel mathematical framework for feature selection in scRNA-seq data.
• To compare the performance of continuous Entropy Sort Feature Weighting (cESFW) against HVG selection.
• To apply cESFW to map early human embryonic development and identify novel cell states and lineage dynamics.

Main Methods:

• Development and application of continuous Entropy Sort Feature Weighting (cESFW) for gene feature selection.
• Analysis of synthetic datasets to benchmark cESFW against HVG selection.
• Application of cESFW to six merged scRNA-seq datasets of human early embryo development without data smoothing or augmentation.

Main Results:

• cESFW effectively distinguishes cell-state-specific genes, outperforming HVG selection on synthetic data.
• Application to human embryonic development data generated a high-resolution embedding revealing 15 distinct cell states and coherent developmental progression.
• Identified previously obscured branch point populations, including morula cells specifying inner cell mass or trophectoderm, and quantified relationships between stem cell cultures and embryonic cell types.

Conclusions:

• Entropy Sorting offers a powerful alternative to HVG selection for scRNA-seq data analysis.
• cESFW accurately captures complex developmental trajectories and identifies novel cell states in early human embryogenesis.
• The method provides valuable insights into lineage decisions, pluripotency, and gene expression dynamics during embryonic development.