The source code of DELPHI web server

The DELPHI source code is designed for high-throughput prediction. It does not have the limitation of 10 sequences per run. We recommend to use the web version of DELPHI if you input is small.

Citation

Yiwei Li, G. Brian Golding, and Lucian Ilie, DELPHI: accurate deep ensemble model for protein interaction sites prediction. Bioinformatics, 2020, btaa750

Contact:

Yiwei Li ([email protected])

Brian Golding ([email protected])

Lucian Ilie ([email protected])

system requirement

DELPHI is developed under Linux environment with python 3.5. Recommended RAM: > 24GB. The RAM requirement mainly depends on the length of the input sequence.

installation

1. clone the source code of DELPHI

mkdir -p Src && cd Src
git clone [DELPHI git link]

2. install python packages. Python virtual environment or conda is recommended for package management.

For GPU version:
pip3 install -r requirement_gpu.txt

For CPU version: 
pip3 install -r requirement_cpu.txt

3. install dependencies

create a program directory
mkdir -p ../programs && cd ../program

• install SPRINT

git clone https://github.com/lucian-ilie/SPRINT.git
git checkout DELPHI_Server
make compute_HSPs_parallel

• install psiblast: 2.6.0+ and download the corresponding nr database. The database is large. You computing cluster should probaly already have a local copy of it.

For Ubuntu:
sudo apt-get install ncbi-blast+

• intall hh-suite. The database used in DELPHI is uniprot20_2015_06.

• intall GENN+ASAquick

• install ANCHOR

Running DELPHI

./run_DELPHI.sh [input_sequence]

DELPHI model architecture