Parsing reads from SAM/BAM by orientation

[winsettz]

Doing some NGS quality control. I'm mapping reads back to a host genome I know is in there, and I'm looking for reads with anomalous behavior, such as forward-forward or reverse-reverse reads.

I /think/ in SAM flags, this case:

5'___________________3'
3'_<--__________<--___5'

Would be

Read reverse strand/SEQ being reverse complemented (bit 16)
Mate reverse strand/SEQ of the next segment in the template being reversed (bit 32)

Which corresponds to -f 48.

For the

5'___-->_____________-->__3'
3'_______________________5'

case, am I correct in assuming it is -F 48?

And for classic reads, it is either read reverse strand (-f 16) or the mate which is reversed (-f 32).

The upper read in this case is forward strand, and its mate is on the reverse. In the case of the bottom read, its flag would be read-reverse-strand, but not its mate.

5'___-->__________________3'
3'___________________<--__5'


5'___<--__________________3'
3'___________________-->__5'

Do I have it right, or am I confused? I've been digging through the blogs for answers, but haven't found any yet. That said, I'm going to manually extract those reads by flag from samtools and remap to reference and see how they look.

[swbarnes2]

Looks right to me.

[winsettz]

The other obvious duh I forgot: If I want only reads that are paired, everything is +1 bit, so

-f 49, but for the second case which I am calling -F 48, I think samtools can take -f and -F simultaneously, so it would be -f 1 -F 48.

And if I wanted to be sophisticated:
5'___________________3'
3'_<--__________<--___5'
R1=113 (1+16+32+64), or: read-paired, read reverse-strand, mate reverse strand, first-in pair
R2=177 (1+16+32+128), or: read-paired, read reverse-strand, mate reverse strand, first-in pair

And
5'___-->_____________-->__3'
3'_______________________5'
R1= -f 65 (1+64, read-paired, first-in-pair), then -F 48 (*not* read-reverse-strand, *not* mate-reverse-strand).
R1= -f 128 (1+128, read-paired, second-in-pair), then -F 48 (*not* read-reverse-strand, *not* mate-reverse-strand).

Otherwise, the flags will also bring in data that doesn't have the -f 1 bit, and throw in a bunch of unpaired data? Though if one doesn't have the paired bit, then it's unlikely the other bits would be included. I wonder how SAM filters, but I think the filter simply looks for the /inclusion/ of the bits you mark, and treats them as wildcard?

Edit:

------------------------------------
At the end of the day, -f 48 covers the <- <- case.

The -> -> case with -F 48 gave me un-even counts in R1 vs R2, so I went with

-f 1 (paired) -F 60