SRAMP is a popular mammalian m6A site predictor we previously developed (Nucleic Acids Res 2016). SRAMP has been totally cited by more than 570 papers (Google Scholar, 4-16, 2024) and represents the mostly used algorithm in this field. A large number of m6A sites were identified by the helps of SRAMP.
After ~8 years after its development, Now we released deepSRAMP (www.cuilab.cn/deepsramp) , which is designed based on a combined framework of transformer neural network and recurrent neural network by fusing the sequence and genomic position features. The results showed that SRAMP2 greatly outperforms its predecessor SRAMP with 15.0% increase of AUROC and 30.9% increase of AUPRC, and greatly outperforms other state-of-the-art m6A predictors (WHISTLE and DeepPromise) with average 16.1% and 18.3% increase of AUC and 43.9% and 46.4% increase of AUPRC, respectively.
deepSRAMP was updated to new model; Adding local version software;deepSRAMP was released;SRAMP can be accessed here.
School of Sports Medicine, Wuhan Institute of Physical Education, No.461 Luoyu Rd. Wuchang District, WuHan 430079, Hubei Province, China
Department of Biomedical Informatics, State Key Laboratory of Vascular Homeostasis and Remodeling, School of Basic Medical Sciences, Peking University, 38 Xueyuan Rd, Beijing 100191, China
The source code of deepSRAMP can be downloaded from zhfanrui/deepSRAMP from Github.
And the trained model can also be downloaded from Github or as below.
Install conda and create a virtual enviroument named sramp with python installed.
conda create -y -n sramp python &&
conda activate sramp
Clone this repo.
git clone https://github.com/zhfanrui/deepSRAMP.git
Install this package and dependencies.
cd deepSRAMP &&
pip install .
Download BLAST and Muscle.
sh setup_inference.sh
Prepare your query fasta and run
deepsramp predict \
--fasta /path/to/fasta.fa \
--db /path/to/database/hg38_mature \
--blast /path/to/blast/bin \
--model /path/to/model/full_400_mature.model \
--out ./result.csv