PhytoCell is a robust, scalable ML-powered solution for annotating plant cell subpopulations and identifying marker genes, advancing machine learning integration in plant genomics.

Operating without relying on prior biological knowledge, PhytoCell preserves intrinsic data structure with a minimal set of marker genes, keeping annotation fidelity.

The framework is data-driven, reducing dependence on prior knowledge and maintaining data structure even when biomarkers are limited.

The Crop Journal published a March 2026 article on PhytoCell, with March 27, 2026 cited as the publication date.

Broader implications for agricultural biotechnology include enabling high-resolution gene expression mapping to support targeted breeding and genetic engineering for stress tolerance, yield, and quality.

PhytoCell stacks four distinct ML models and uses maximal information coefficient calculations to rank genes and iteratively select marker candidates in a data-driven loop.

The four-model stack enhances predictive stability and generalization by using part of the data to identify biomarkers and then analyzing the remaining data.

PhytoCell recovers conventional marker genes and also uncovers additional genes overlooked by other methods, expanding resources for crop improvement and cellular mechanism studies.

The framework uncovers novel marker genes not detected by conventional pipelines, highlighting its exploratory power for crop improvement and biology.

Open availability and a web server aim to democratize access to advanced ML techniques for plant scRNA-seq analysis.

A user-friendly web server accompanies PhytoCell, enabling researchers with varying computational expertise to explore marker genes and perform automated cell-type annotation.

The freely accessible web server for marker gene exploration and cell-type annotation is available at https://cgris.net/phyto.