ConSurf-debug

Stand Alone version of ConSurf with detailed tutorial - (installation + database + usage)

Start from a Nature article

• OCA-T1 and OCA-T2 are coactivators of POU2F3 in the tuft cell lineage - 7 July, 2022 - Nature

About ConSurf

• https://consurf.tau.ac.il/consurf_index.php
• Stand alone python wrapper: https://consurf.tau.ac.il/STANDALONE/stand_alone_consurf-1.00.rar
• ConSurf DB: https://consurfdb.tau.ac.il/ [not useful, files cannot be downloaded]
• ConSurf Github: https://github.com/Rostlab/ConSurf [not update anymore since 2015. Just take a glimpse, we won't use it]

Installation

1. download python script (wrapper) from https://consurf.tau.ac.il/STANDALONE/stand_alone_consurf-1.00.rar

2. download Rate4Site from https://www.tau.ac.il/~itaymay/cp/rate4site.html

• ftp://rostlab.org/rate4site/

tar zxvf rate4site-3.0.0.tar.gz
cd rate4site-3.0.0
./configure --prefix=/home/YOUR/PATH
make
make install

• Note1: It could be successfully installed, but with some bugs like below. We can just ignore it.
• Note2: The default algorithm -im is broken, we must use -ib, so we will add --Maximum_Likelihood when running python stand_alone_consurf.py.

# error:
rate4site: errorMsg.cpp:41: static void errorMsg::reportError(const string&, int): Assertion `0' failed.
Aborted (core dumped)

3. create conda env [see instructions.txt file]

# [official suggestion] Load modules: module load python/python-3.8 hmmr/hmmr-3.1b2 clustalw/2.1
conda create --name consurf python=3.8
conda activate consurf
conda install -c bioconda hmmer=3.1 clustalw=2.1 cd-hit mafft=7 prank muscle=3
# additional python packages
pip install biopython requests

4. download the databases you need

# I just need this, 67G after decompression
wget -b https://ftp.uniprot.org/pub/databases/uniprot/uniref/uniref90/uniref90.fasta.gz
gzip -d uniref90.fasta.gz

4. modify configuration file of GENERAL_CONSTANTS.py: the path of all tools and databases are stored in this file.

CD_HIT_DIR = "/home/zz950/softwares/miniconda3/envs/consurf/bin/cd-hit"
UNIREF90_DB_FASTA = "/home/zz950/softwares/ConSurf/db/uniref90.fasta"

some tool path need to be modified in python script, stand_alone_consurf.py

PRT_JAR_FILE = "/home/zz950/softwares/ConSurf/prottest-3.4.2/prottest-3.4.2.jar"

5. Finally, run the main python script stand_alone_consurf.py

# these are the minimun parameters, don't remove any of them.
python stand_alone_consurf.py --algorithm HMMER --Maximum_Likelihood --seq /home/zz950/softwares/ConSurf/test.fasta --dir /home/zz950/softwares/ConSurf/test

6. Compare results between stand alone and web server versions. Confirmed. The same.

Others

• all log info is in test/log.txt file. It contains the full commands and error info. It's useful for debug.
• a similar tool using deep learning. vespa_emb Input_protein_seq -o data/embeddings.h5 --prott5_weights_cache data/cache [don't like it!]
• One module (-im) of rate4site is broken. Likelihood after optimization is -0x1.c39f46ba773c5p+14. The log(likelihoodsV[pos]) is negative and the script will stop. So we will use (-ib) module [These are rate inference methods].

Assertion failed: (log(likelihoodsV[pos])>0.0), function computeML_siteSpecificRate, file siteSpecificRate.cpp, line 27.
Abort trap: 6

How I debug?

• see log.txt file, check where the program stopped;
• use sublime, search the source code folder to locate the code that send the error;
• insert print or log code to output the logic of the program;
• soon, you will find the bug.