sorted bam larger than unsorted bam

[ians]

Recently i've sorted 2 different alignments and in both cases the sorted bam is ~3x larger in disk size. I did the sort twice with the same result.
Does anyone have a suspicion of what is going on here?

[xied75]

1, output bam uncompressed? (show the command line pls.)
2, you are sorting on name or coord? could it be that bgzip block can't compress that hard after the sort? (Very unlikely though.)

[ians]

Here's what i'm running:

time -v samtools sort $mergedBam $sortedBam

Another run yielded the same result.

[jkbonfield]

Do you have a combination of shallow coverage and excessively long read names? That can cause the sorting to be of little benefit to sequence and positional compression while also being detrimental to read name compression.

[ians]

headers aren't anything out of the ordinary. Here are a few entries from the sorted bam:

HWI-ST1063_0137:6:1308:10615:65342#0	161	chr1	765	59	95M6S	=	1019	354	TGACGGACTACATGAGATAGAAGAGAGAATTTTGGGAGCAGAAGATATCATAGAAAACATTGACACAACCTTCAAAGAGAACGTAAATAGGAAAAAGCTCC	?=++4B0@FHDB:<+C::CE@HC>?38C9CHAFHGEDHIJCDAFCHGI@FCGBH@@AFEGIIEEHHGH??DFD@>@CA3;?A>A;<<BAC:@CDD######	PG:Z:novoalign	AS:i:3UQ:i:36	NM:i:0	MD:Z:95
HWI-ST1063_0137:6:1308:10615:65342#0	81	chr1	1019	1	101M	=	765	-354	ATAAATGTCCATAAGTAGACATGAAGCCTGCAGAATTCCAAATAGAATGGACCAGAAAATAAATTCCTCCTGTCACATAATAGTCAAAACACCAAATGCAC	>>;>;5;@;;A;7.))7;7@C=?CA;@DAA=ADC;ECC=DB@IED<DBB?0D99??0499:<9DDC88:2<F9FAFDCA?CCBEE<:C<?DDABDB+??1?	PG:Z:novoalign	AS:i:1UQ:i:14	NM:i:1	MD:Z:14A86	CC:Z:=	CP:i:71209	ZS:Z:R	ZN:i:3	NH:i:3	HI:i:1	IH:i:3
HWI-ST1063_0137:6:2303:18332:58082#0	419	chr1	1044	1	13S88M	=	1151	210	GAACATAGAAGAAGCCTGCAGAATTCCAAATAGAATGGACCAGAAAATAAATTCCTCCTGTCACATAATAGTCAAAACACCAAATGCACAAAACAAAGAAT	CCCFFFFFHHHHHJJJJJJJGGJJJJJJIGIGJJJIJIIJJJJJJJJIHIEHIDIGIIIIGIGGIJG@GGGGIIIJJHHHHFFFEAACCEEBDDDDACDDC	PG:Z:novoalign	AS:i:105	UQ:i:105	NM:i:0	MD:Z:88	PQ:i:135	SM:i:70	AM:i:70	ZS:Z:R	ZN:i:3	NH:i:3	HI:i:3	IH:i:3

This is an RNAseq project and we have sufficient number of reads to get a deep sampling. I suspect something is going wrong at this step because down stream I am getting 20% coverage over the genome with this sample...


I aligned using novoalign, if that is of any consequence.