Based on the manual, featureCounts function in Rsubread set countPrimaryAlignmentsOnly=FALSE by default. By setting countPrimaryAlignmentsOnly=TRUE, as alignments used are less, the counts seem to be less too. But I got more counts in all samples. Sometimes much more. For example, I got this when not setting countPrimaryAlignmentsOnly (=FALSE)

ENSG00000227232.5     1       1       2       2       1       1

But when I set countPrimaryAlignmentsOnly=TRUE, I got this

ENSG00000227232.5     190     217     210     256     235     254

I used Rsubread 1.18.0 in R 3.2.1. The samples are RNA single-end data. Mapper was STAR 2.5.1b. Other settings of featureCounts() were: isGTFAnnotationFile=TRUE, nthreads=16, strandSpecific=2. All other settings are default. I used an external human GTF file from GENCODE (v24, with patches and scaffolds).

The code is as below. Did I do anything wrong? I'm glad to provide further info if needed. Thanks.

library(Rsubread)
fc <- featureCounts(files=c('G1/G1.STAR.genome.bam', 'G2/G2.STAR.genome.bam', 'G3/G3.STAR.genome.bam', 's1/s1.STAR.genome.bam', 's2/s2.STAR.genome.bam', 's3/s3.STAR.genome.bam'), countPrimaryAlignmentsOnly=TRUE, annot.ext='gencode.v24.chr_patch_hapl_scaff.annotation.gtf', isGTFAnnotationFile=TRUE, nthreads=16, strandSpecific=2)
colnames(fc$counts) <- c('G1', 'G2', 'G3', 's1', 's2', 's3')
write.table(fc$counts, file='count.txt', quote=FALSE, sep='\t', col.names=TRUE, row.names=TRUE)

> sessionInfo()
R version 3.2.1 (2015-06-18)
Platform: x86_64-unknown-linux-gnu (64-bit)
Running under: CentOS release 6.4 (Final)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
 [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] Rsubread_1.18.0

loaded via a namespace (and not attached):
[1] tools_3.2.1

When 'primaryOnly' option is tuned on, those multi-mapping reads that were marked as 'primary alignment' were counted. However, no multi-mapping reads were counted when 'primaryOnly' option is tuned off. That is the reason why you got more counts when you turned on this option.

The help page for this option states "If TRUE, all primary alignments in a dataset will be counted no matter they are from multi-mapping reads or not (ie. countMultiMappingReads is ignored)."

For instance, your bam file "G1.G1.STAR.genome.bam" includes 32992731 multi-mapping reads, of which 26164591 are marked as 'secondary alignment'. No multi-mapping reads were considered when 'primaryOnly' option is turned off, however ~7 million (32992731 - 26164591) multi-mapping reads that were reported as primary alignments were considered in counting when this option is tuned on.

Firstly please update your Rsubread version to its latest (1.20.3). The parameter 'countPrimaryAlignmentsOnly' is now called 'primaryOnly'.

Could you provide summary stats for your runnings with or without this parameter? You can do so by running the following code:

library(Rsubread)
fc1 <- featureCounts(files=c('G1/G1.STAR.genome.bam', 'G2/G2.STAR.genome.bam', 'G3/G3.STAR.genome.bam', 's1/s1.STAR.genome.bam', 's2/s2.STAR.genome.bam', 's3/s3.STAR.genome.bam'), countPrimaryAlignmentsOnly=TRUE, annot.ext='gencode.v24.chr_patch_hapl_scaff.annotation.gtf', isGTFAnnotationFile=TRUE, nthreads=16, strandSpecific=2)
fc1$stat
fc2 <- featureCounts(files=c('G1/G1.STAR.genome.bam', 'G2/G2.STAR.genome.bam', 'G3/G3.STAR.genome.bam', 's1/s1.STAR.genome.bam', 's2/s2.STAR.genome.bam', 's3/s3.STAR.genome.bam'), countPrimaryAlignmentsOnly=FALSE, annot.ext='gencode.v24.chr_patch_hapl_scaff.annotation.gtf', isGTFAnnotationFile=TRUE, nthreads=16, strandSpecific=2)
fc2$stat