Arabidopsis proteome annotation

Protein localization,resolved for Arabidopsis.

AtSubP-2.0 uses an Arabidopsis-specific convolutional neural network to predict where proteins act within the cell—from a single FASTA sequence to a four-level localization report.

Run a prediction Standalone package

Prediction hierarchy

One sequence, four decisions

Level I

Single vs dual

Primary routing

Level II

12 classes

Single localization

Level III

9 pairs

Dual localization

Level IV

Membrane type

Single- or multi-pass

From sequence to compartment

The complete prediction workflow

TPC · 8,000 features · CNN

AtSubP-2.0 workflow from Arabidopsis protein sequence and tripeptide composition through a convolutional neural network and four-level localization hierarchy — AtSubP-2.0 four-level protein subcellular localization workflow.Generic query rows illustrate the report format.

Why species-specific?

Built around the Arabidopsis proteome

Arabidopsis thaliana is a foundational model for plant biology. Knowing where its proteins localize helps researchers interpret trafficking, interactions, regulation, and organelle-specific function.

AtSubP-2.0 converts each sequence into an 8,000-dimensional tripeptide-composition vector and applies CNN models across four linked localization levels.

Publication

Please cite AtSubP-2.0

Duhan, N., & Kaundal, R. (2025). AtSubP-2.0: An integrated web server for the annotation of Arabidopsis proteome subcellular localization using deep learning. The Plant Genome, 18(1), e20536.DOI: 10.1002/tpg2.20536