MuTect2 does not call tumor-only variants
I'm trying to use MuTect2, however I'm not getting tumor-specific variants called. To get MuTect2 working I'm using an artificial tumor/normal data set, however, somatic mutations I introduced are not detected.
I use an arbitrarily selected 100 kb region of chromosome 19 and two sets of variants from the Genome-in-a-Bottle database to generate the tumor and normal bam's (using bamsurgeon) whereby the tumor batch of 67 SNPs contains also all 33 'normal' SNPs All SNPs have allele frequencies close to 1.
For variant calling I followed in general the Best Practice pipeline used the picardTools MarkDuplicates and recalibrated and then called MuTect2 with
java -jar /usr/local/src/GenomeAnalysisTK-3.6/GenomeAnalysisTK.jar -T MuTect2 -R /data/references/human/genome/GRCh37_hg19/hg19_wochrs.fasta -I:tumor reg5_T_67snps_noDup_wRG_recal.bam -I:normal reg5_N_33snps_noDup_wRG_recal.bam -o reg5_redSNPs_MuTect2_redetection.vcf -L Regions_files/reg5.bed -dontTrimActiveRegions -forceActive
However, the tumor-only sites are not correctly called. The output states "0 active regions".
On the other hand, when I perform MuTect2 calling only with a single sample (as "-I:tumor") applying "--artifact_detection_mode" all variants are detected correctly. Likewise, the HaplotypeCaller calls the variants correctly for both sets.
Any thoughts? What am I missing when calling on the tumor/normal pair?