All small python scripts require environment.yml to be installed to work.
Prepare a list of identified SMILES to fragment. For the moment, we take the structures from https://doi.org/10.5281/zenodo.6378223 as starting point. To prepare the structures to fragment, just run lotus2cfm.py:
python scripts/lotus2cfm.py
This installation procedure works on the UniGE HPC. This does not mean it will work on another.
First, create a cmf-4 directory:
mkdir cfm-4
Then load requested modules:
module load GCC/6.3.0-2.27 Singularity/2.4.2
Build cfm-4 from its last Docker image:
singularity build cfm-4/cfm.sif docker://wishartlab/cfmid
Pull the previsouly generated smiles list (this command is not generic, it needs to be adapted):
scp smiles4cfm.txt [email protected]:smiles.txt
Also pull last bash commands:
scp scripts/run_cfm_test.sh [email protected]:run_cfm_test.sh
scp scripts/run_cfm.sh [email protected]:run_cfm.sh
scp scripts/run_cfm_neg.sh [email protected]:run_cfm_neg.sh
Create a test file with 10 structures to check if everything works fine:
head smiles.txt -n 10 > test.txt
Create a test directory:
mkdir test
Split the test file asit would be split if real:
split --lines=1 --numeric-suffixes=1 --suffix-length=4 --additional-suffix=.txt test.txt test/test-
Create a testout directory:
mkdir testout
Run run_cfm_test.sh in a sbatch array:
sbatch --array=1-10 run_cfm_test.sh
Depending on the length of your SMILES list, you may want to split it:
First, create a smiles directory:
mkdir smiles
Split the big file:
split --lines=100 --numeric-suffixes=1 --suffix-length=4 --additional-suffix=.txt smiles.txt smiles/smiles-
Count how many entries it generated for the next step
ls smiles/ | wc -l
Create the posout directory:
mkdir posout
Run run_cfm.sh in a sbatch array (adapt the array length):
sbatch --array=1-2875 run_cfm.sh
Create the negout directory:
mkdir negout
Run run_cfm_neg.sh in a sbatch array (adapt the array length):
sbatch --array=1-2875 run_cfm_neg.sh
Download CFM fragmentation results from the baobab server (this command is not generic, it needs to be adapted): (We first zip them before for an efficient transfer)
zip -r results.zip ./posout
zip -r results_neg.zip ./negout
scp [email protected]:results.zip ./results.zip
scp [email protected]:results_neg.zip ./results_neg.zip
unzip results.zip
unzip results_neg.zip
The output of cfm-predict consist of .log file containing mass spectra, where each fragments are individually labelled and eventually linked to a substrcture. Such information might be usefull later but for now we only want to keep the raw ms data: If you want to merge the three different energies you can choose between 'max','mean', and 'sum' for the moment.
python scripts/log2mgf.py results/ log sum
python scripts/log2mgf.py results_neg/ log sum
We need to prepare and adducted table containing the protonated and deprotonated masses, run prepare_headers.py:
python scripts/prepare_headers.py smiles4cfm.txt s+ adducted.tsv YOUR_SMILES_COLUMN_NAME YOUR_SHORT_INCHIKEY_COLUMN_NAME
We can now populate each mgf with its corresponding metadata, run populate_headers.py:
python scripts/populate_headers.py adducted.tsv results/ positive
python scripts/populate_headers.py adducted.tsv results_neg/ negative
We concatenate each documented mgf files to a single spectral mgf file, run concat.sh:
bash scripts/concat.sh ./results isdb_pos.mgf
bash scripts/concat.sh ./results_neg isdb_neg.mgf
For multiple reasons, some entries might not be fragmented. We here list the ones correctly fragmented to facilitate further investigation.
find ./results -type f -name '*.mgf' | sed 's!.*/!!' | sed 's!.mgf!!' > list_fragmented_pos.txt
find ./results_neg -type f -name '*.mgf' | sed 's!.*/!!' | sed 's!.mgf!!' > list_fragmented_neg.txt
Can be found at: