Different number of total reads in different alignments using the same original readset

Question

[malteherold]

I have 3 Assemblies (MT, MG and Combined) and I mapped 2 different sets of reads (MT / MG) to each of these assemblies. I noticed that the number of "total reads" that is reported for the resulting bam files varies, even though I originally used the same number of reads.

For the mapping I used BWA mem for paired reads and for singletons and then merged the 2 resulting files, before sorting and indexing:

samtools merge -@ ${THREADS} -f $PREFIX.merged.bam \
  <(bwa mem -v 1 -t ${THREADS} -M -R \"$SAMHEADER\" ${ASSEMBLY} ${PE1} ${PE2} | \
  samtools view -@ ${THREADS} -bS -) \
  <(bwa mem -v 1 -t ${THREADS} -M -R \"$SAMHEADER\" ${ASSEMBLY} ${SE} | \
  samtools view -@ ${THREADS} -bS -)

I counted the reads in the resulting bam files (also with flagstat and idxstats) and got the following numbers:

parallel "echo {};samtools view -c {}" ::: *.bam

MT reads in fastq files combined: 3236023

MT_vs_MT.sorted.bam
3236025
MT_vs_MG.sorted.bam
3236593
MT_vs_Combined.sorted.bam
3236408

MG reads in fastq files combined: 19920596

MG_vs_Combined.sorted.bam
19961083
MG_vs_MT.sorted.bam
19920896
MG_vs_MG.sorted.bam
19951246

Why is the number of total reads in the alignment files different to the original number and between the files?

Thank you in advance.

[Devon Ryan]

How many are marked as secondary and or supplementary?

[WouterDeCoster]

Indeed, some reads can map equally well to multiple locations, causing this discrepancy.

[malteherold]

supplementary: samtools view -c -f 2048 : 0 everywhere

non primary: samtools view -c -f 256

MT_vs_MT.sorted.bam
2    
MT_vs_MG
570
MT_vs_Combined
385  

MG_vs_Combined
40487
MG_vs_MT
300
MG_vs_MG
30650

And yes these numbers add up to the total read numbers... Thanks for your answer. How do I close this or mark as accepted?