SynthDNM is a random-forest based classifier that can be readily adapted to new sequencing or variant-calling pipelines by applying a flexible approach to constructing simulated training examples from real data. The optimized SynthDNM classifiers predict de novo SNPs and indels with robust accuracy across multiple methods of variant calling.

References:

Aojie Lian, James Guevara, Kun Xia and Jonathan Sebat
Customized de novo mutation detection for any variant calling pipeline: SynthDNM
Bioinformatics, Volume 37, Issue 20, October 2021, Pages 3640–3641, https://doi.org/10.1093/bioinformatics/btab225

• SynthDNM Github page

• Module Name: SynthDNM (see the modules page for more information)
• Unusual environment variables set
  • SDNM_HOME  installation directory
  • SDNM_BIN       executable directory
  • SDNM_DATA    sample data folder

Allocate an interactive session and run the program.
Sample session on a GPU node:

[user@biowulf ~]$ sinteractive --mem=8g -c4 --gres=lscratch:20
[user@cn2379 ~]$ module load synthdnm
[+] Loading singularity  3.8.4  on cn2379
[+] Loading synthdnm  1.1.50

[user@cn2379 ~]$ run_synthdnm.py -h
usage: run_synthdnm.py [-h] [--vcf_file VCF_FILE] --ped_file PED_FILE
                       [--region REGION] [--features_file FEATURES_FILE]
                       [--output_folder OUTPUT_FOLDER]
                       [--training_set_tsv TRAINING_SET_TSV]
                       {classify,make_training_set,train,grid_search} ...

SynthDNM: a de novo mutation classifier and training paradigm

positional arguments:
  {classify,make_training_set,train,grid_search}
                        Available sub-commands
    classify            Classify DNMs using pre-trained classifiers.
    make_training_set   Make training set.
    train               Train classifiers
    grid_search         Randomized grid search across hyperparameters.

optional arguments:
  -h, --help            show this help message and exit
  --vcf_file VCF_FILE   VCF file input
  --ped_file PED_FILE   Pedigree file (.fam/.ped/.psam) input
  --region REGION       Interval ('{}' or '{}:{}-{}' in format of chr or
                        chr:start-end) on which to run training or
                        classification
  --features_file FEATURES_FILE
                        Features file input
  --output_folder OUTPUT_FOLDER
                        Output folder for output files (if not used, then
                        output folder is set to 'synthdnm_output')
  --training_set_tsv TRAINING_SET_TSV
                        Training set file (created using make_training_set
                        mode)
[user@cn2379 ~]$ run_synthdnm.py classify -h
usage: run_synthdnm.py classify [-h] --clf_folder CLF_FOLDER
                                [-feature_extraction_only]

optional arguments:
  -h, --help            show this help message and exit
  --clf_folder CLF_FOLDER
                        Folder that contains the classifiers, which must be in
                        .pkl format (if not specified, will look for them in
                        the default data folder)
  -feature_extraction_only
                        Only output the features file (without classifying

[user@cn2379 ~]$ exit

Create a batch input file (e.g. synthdnm.sh). For example:

#!/bin/bash
set -e
module load synthdnm
cp $SDNM_DATA/* .
synthdnm -v tutorial.vcf -f tutorial.ped

Submit this job using the Slurm sbatch command.

sbatch synthdnm.sh