bgm-cwg / novocaller

novoCaller

About

novoCaller is a Bayesian de novo variant calling algorithm that uses information from read-level data both in the pedigree and in unrelated samples. The method was extensively tested using large trio sequencing studies, and it consistently achieved over 98% sensitivity while giving significantly more specificity than other well known methods for the same sensitivity values.
The method works with a vcf file alone (first layer), but it can also inspect the reads in the bam file and compute probabilities based on direction-specific (forward and reverse) reads (second layer).
We made the software keeping VCFv4.1 in mind. The software looks at the format field of each variant call location in the VCF file and reads the AD element. Then it extracts that part from every sample.

Compile:

g++ novoCaller.cpp -o novoCaller

Usage:

Run first layer (C++ code) - uses only VCF:

./novoCaller \
-I <path to vcf file> \
-O <path to output file for layer 1> \
-T <path to file containing sample IDs of the trios, the IDs are in the order:parent1(TAB)parent2(TAB)proband> \
-X <put 1 if you want to run on X chromosome as well, 0 otherwise> \
-P <threshold on posterior probability. Calls are made if the PP is above threshold. Use a low value like 0.005 so that a large number of calls are made for the second layer> \
-E <threshold on the ExAC allele frequency, e.g. 0.0001>

Run second layer (Python code) - uses BAM files (OPTIONAL):

python -W ignore novoCallerBAM.py \  
-I <path to the output file from previous step (the file given in -O option)> \
-U <path to a file containing paths to the bam files from unrelated samples> \
-T <path to a file containing paths to the bam files of the trio> \
-O <path to the output file for the second layer>

The ignore option is given to ignore log of 0 warning.

Example command line:

./novoCaller -I ./all_calls.vep.vcf -O step1_out.txt -T trio_ids.txt -X 1 -P 0.005 -E 0.008 
python -W ignore novoCallerBAM.py  -I step1_out.txt -U de_novo_unrelated_bams.txt -T de_novo_case_bams.txt -O denovo_calls.txt 

Output format:

denovo_calls.txt columns are the following: Rank, chromosome, position, reference allele, alternative allele, AF (in samples), rhos, priors, PP=posterior probability, AF_unrelated, gene_name(s)

References

Find our paper here

novocaller's Issues

Providing example files

Hi developers of novoCaller,

I have tried running the first layer of novoCaller with the following command but the program just keep on running for over 24 hours without generating any output data. I am new to bioinformatics so please correct me if I made any mistakes.

Command:
novoCaller -I input.vcf -O step_1_out.txt -T sample_id.txt -X 1 -P 0.005 -E 0.008

vcf:
example.vcf.gz

sample ID file:
sample_id.txt

It would be very helpful if you can provide example files for the program.

Thanks a lot!

Marcus

Pysam error when running novoCallerBAM.py

Hi,

When trying to run the python script using the bam files and the output of step one, I get this error:
Traceback (most recent call last):
File "/nfs/projects/refractory_epilepsy/Novocaller_test/novoCaller/novoCallerBAM.py", line 802, in
runner(outfilename,initial_filename,unrelated_filename,trio_filename)
File "/nfs/projects/refractory_epilepsy/Novocaller_test/novoCaller/novoCallerBAM.py", line 731, in runner
PP,ADfs,ADrs,ADfs_U,ADrs_U,rho_f_new,rho_r_new,prior_L_new,AF_unrel = PP_calc(trio_samfiles,unrelated_samfiles,chrom,pos,REF,ALT,allele_freq,MQ_thresh,BQ_thresh)
File "/nfs/projects/refractory_epilepsy/Novocaller_test/novoCaller/novoCallerBAM.py", line 628, in PP_calc
ADfs,ADrs = get_all_ADs_combined(unrelated_samfiles,chrom,pos,REF,ALT,MQ_thresh,BQ_thresh)
File "/nfs/projects/refractory_epilepsy/Novocaller_test/novoCaller/novoCallerBAM.py", line 312, in get_all_ADs_combined
ADf,ADr = get_ADs_combined(samfile,chrom,position_actual,REF,ALT,MQ_thresh,BQ_thresh)
File "/nfs/projects/refractory_epilepsy/Novocaller_test/novoCaller/novoCallerBAM.py", line 303, in get_ADs_combined
ADf,ADr = get_ADs(samfile,chrom,position_actual,REF[0],MQ_thresh,BQ_thresh)
File "/nfs/projects/refractory_epilepsy/Novocaller_test/novoCaller/novoCallerBAM.py", line 149, in get_ADs
SP=samfile.pileup("chr"+CC, position, position+1)
File "pysam/libcalignmentfile.pyx", line 1326, in pysam.libcalignmentfile.AlignmentFile.pileup
File "pysam/libchtslib.pyx", line 685, in pysam.libchtslib.HTSFile.parse_region
ValueError: invalid contig chr1

unrelated file

seems I put this question on the wrong page. sorry, I should put it at here; https://github.com/dbmi-bgm/granite, but I cannot find the issues button on the granite page...

Hi, i have some questions when i using novocaller, maybe someone can help me figure out some of these:

1. Can novo caller detect de novo indel?
2. when calling DNVs, the unrelated files refer to a) independent trios from the specific trios; b) samples with unrelated phenotypes that not suppose to carry DNV in this position; 3) samples that are independent of each other and the specific trios.
3. how many un-related samples will novoCaller prefer
4. we know sequencing error and gatk may detect more than one DNVs in many trios at an identical position, which is not a real DNV but a sequencing error, in case I used many other trios as un-related samples and the error DNVs in other trios at this position will increase or decrease the DNV probability
5. I saw the example vcf file, is it necessary to put the child at the first genotype column [10th column in vcf file]. the reason I ask this question is because I have many trios, if it is necessary to put the child in the 10th column in vcf file, that means I have to create many big vcf files containing similar un-related samples for each trio