lib_format_counts.json

[Daniel Hyduke]

I've run salmon on unstranded paired end, illumina data with the and have been getting warnings like:
  "Detected a *potential* strand bias > 1% in an unstranded protocol check the file SALMON/lib_format_counts.json for details"
and, I was wondering if everything looks okay in the lib_format_counts.json file?  Is there a reason that IU has a value of 0 and ISF is so large?

salmon quant -l IU -i k_31 -1 *R1.fastq.gz -2 *R2.fastq.qz -o SALMON --gcBias --seqBias -p 49 --numBootstraps 100 

{ ...,

   "expected_format": "IU",
    "compatible_fragment_ratio": 0.99785717440557498,
    "num_compatible_fragments": 60999508,
    "num_assigned_fragments": 61130500,
    "num_consistent_mappings": 140054768,
    "num_inconsistent_mappings": 4642302,
    "strand_mapping_bias": 0.061818987840528215,
    "MSF": 0,
    "OSF": 206245,
    "ISF": 131396724,
    "MSR": 0,
    "OSR": 19754,
    "ISR": 8658044,
    "SF": 3504476,
    "SR": 908434,
    "MU": 0,
    "OU": 0,
    "IU": 0,
    "U": 0
}

[Rob]

Hi Daniel,

  The count of 0 for IU is OK.  Specifically, a given fragment will never, itself, be assigned a library type of IU, because a fragment mapping

displays either the "ISF" or "ISR" orientation (i.e. either the first read comes from the forward strand or the reverse strand).  However, what 

I can tell from your lib_format_counts.json log is that your data definitely appears to be stranded.  Typically, in an unstranded (IU) library, you'd

expect ISF and ISR to be present in roughly equal proportions — and you'd expect the "strand_mapping_bias" to be close to 0.5.  However, 

here what we can see is that the vast majority of your mappings look like ISF.  In fact, roughly 94% of the data is consistent with a stranded 

library.  Are you certain this data is unstranded?

Best,

Rob

P.S.  Did you ever resolve the density plot issue you mentioned in your previous post?