drutai

Drutai is a deep learning–based framework for predicting interactions between small molecule drugs and protein targets.
And this is a high-performance drug-target interaction prediction tool
using ONNX models, with easier installation, higher performance as well as less dependency requirements . It supports multiple deep learning architectures including
LSTMCNN, CNN, ConvMixer64, DSConv, MobileNetV2, ResNet18, and SEResNet.

Usage (single-table mode, recommended):
drutai.predict -m lstmcnn \

 -i input.tsv \

 -t fastadir/ \

 -o output.drutai

Usage (legacy mode, compatible with original drutai format):
drutai.predict -m lstmcnn \

 --br relations.txt \

 --smile smiles.txt \

 -t fastadir/ \

 -o output.drutai

Arguments:
-m MODEL Model name (default and recommended: lstmcnn)
-i INPUT Single-table TSV input (columns: sm, target, smile)
--br FILE [Legacy] Drug-protein relation file (columns: sm, target)
--smile FILE [Legacy] Drug SMILES file (columns: sm, smile)
-t FASTADIR Directory containing per-target .fasta files
-o OUTPUT Output file (default: predictionsonnx.drutai)
-j THREADS Number of threads for feature extraction (default: all CPUs)

Available models: lstmcnn, cnn, convmixer64, dsconv, mobilenetv2,

               resnetprea18tf2, scaresnet