|
DeepKOALA is a high-performance deep learning-based tool for rapid protein
function annotation according to the KEGG Orthology (KO) system.
By framing KO assignment as an open-set recognition problem, it can effectively
distinguish between known and unknown functional sequences, thereby reducing
false-positive annotations.
Built on a Gated Recurrent Unit (GRU) architecture, the tool provides excellent
computational efficiency while ensuring high accuracy.
This DeepKOALA runs on CPU machines.
|
|
|
|
|
|
| Enter FASTA Sequences |
|
|
|
| or upload a sequence file |
|
The file size of up to 300 MB with the limit of 500,000 sequences may be uploaded.
|
|
Model
|
|
High-precision annotation for only full-length sequences.
For full-length and fragmented sequences (e.g., choice for metagenome).
|
|
| E-mail |
|
|
|
|
|
|
|
Current database version
|
Download DeepKOALA
|
Reference
-
Yu Z., Meng L., Nguyen C.H., Mamitsuka H., Kanehisa M., OgataH.
DeepKOALA: A Fast and Accurate Deep Learning Framework for KEGG Orthology Assignment.
bioRxiv 2026.01.07.698072; doi:10.64898/2026.01.07.698072
|
