Is it the sample or the software that is wrong?

Hi everyone,

I am conducting a differential expression analysis using a public sra dataset(PRJNA762935, PMID:35157052).Of these, I will discuss SRR15881954~SRR15881986 in particular here.This dataset appears to have been generated using a paired-end stranded protocol.

Refarence transcriptome data was obtained from GENCODE Human Release 30 (GRCh38.p12) as following:

wget ftp://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_human/release_30/gencode.v30.transcripts.fa.gz

I generated an index from this file (salmon index -k 31) and quantified the FASTQ files using the following sample code:

tools/salmon-latest_linux_x86_64/bin/salmon quant -i data/transcripts_index_salmon -p 4 -l A -1 SRR15881954_1.fastq.gz -2 SRR15881954_2.fastq.gz --validateMappings -o SRR15881954_exp_salmon

Most of the cases were guessed as "ISR," but in two cases(SRR15881985, SRR15881986), "IU" was inferred.Here is the output of the cases inferred to be an IU:

>SRR15881985
{
"read_files": [
    "SRR15881985_1.fastq.gz",
    "SRR15881985_2.fastq.gz"
],
"expected_format": "IU",
"compatible_fragment_ratio": 1.0,
"num_compatible_fragments": 14061874,
"num_assigned_fragments": 14061874,
"num_frags_with_concordant_consistent_mappings": 12199998,
"num_frags_with_inconsistent_or_orphan_mappings": 2103817,
"strand_mapping_bias": 0.5223949217040855,
"MSF": 0,
"OSF": 0,
"ISF": 5826781,
"MSR": 0,
"OSR": 0,
"ISR": 6373217,
"SF": 1337245,
"SR": 766572,
"MU": 0,
"OU": 0,
"IU": 0,
"U": 0
}

>SRR15881986
{
"read_files": [
    "SRR15881986_1.fastq.gz",
    "SRR15881986_2.fastq.gz"
],
"expected_format": "IU",
"compatible_fragment_ratio": 1.0,
"num_compatible_fragments": 15302937,
"num_assigned_fragments": 15302937,
"num_frags_with_concordant_consistent_mappings": 13708761,
"num_frags_with_inconsistent_or_orphan_mappings": 1689014,
"strand_mapping_bias": 0.6453805708626768,
"MSF": 0,
"OSF": 0,
"ISF": 4861393,
"MSR": 0,
"OSR": 0,
"ISR": 8847368,
"SF": 1198881,
"SR": 490133,
"MU": 0,
"OU": 0,
"IU": 0,
"U": 0
}

To experiment with an alternate tool to infer library format, I aligned the FASTQ files using HIAST2 and used RSeQC's infer_experiment.py python module using intronic and exonic features per gene in .bed format as input (http://rseqc.sourceforge.net/).In one case(SRR15881985), it was inferred to be an ISR, but in the other case(SRR15881986), RSeQC made an inference that it could be an IU.

>rseqc.guess
SRR15881985
This is PairEnd Data
Fraction of reads failed to determine: 0.0223
Fraction of reads explained by "1++,1--,2+-,2-+": 0.0372
Fraction of reads explained by "1+-,1-+,2++,2--": 0.9405

SRR15881986
Fraction of reads failed to determine: 0.0164
Fraction of reads explained by "1++,1--,2+-,2-+": 0.2993
Fraction of reads explained by "1+-,1-+,2++,2--": 0.6843

From this result, SRR15881985 quantifies by ISR specification in salmon, but how should SRR15881986 specify the library type? Or does the case need some processing?

salmon rseqc rnaseq strand

• 1,782 views • link

written 3.8 years ago by kuwasenai_vegetable •

If these samples are part of the same dataset then it is less likely that they have been processed with two separate kits/methods.

Looks like there is a paper associated with this dataset. I suggest that you look the methods section up:

https://pubmed.ncbi.nlm.nih.gov/35157052/

• 0 views • link

written 3.8 years ago by GenoMax

thank you for your reply !

I have read the paper and it says the following:
"RNA sequencing was performed with Illumina HiSeq 2000 at the Baylor Sequencing Center following standard protocols (61)."
and in reference 61:
"For the Illumina (IL) library preparations, total RNA was first extracted from collected tissues using the Plant RNeasy mini kit (Qiagen) and libraries prepared using mRNA-Seq 8 sample prep kit (Illumina Inc. San Diego, USA, Cat. # RS-100-0801) according to the manufacturer’s protocol.For HTR library preparations, mRNA isolation with both Dynabeads (Invitrogen) and Sera-Mag oligo dT magnetic beads (Thermo Scientific, Cat. # 3815-2103-010150) were performed based on the Dynabeads mRNA direct kit (Invitrogen) protocol with minor adjustments (see Methods 1 in Supplementary Material). The control libraries C1 and C2 were made with total RNA extracted using the Plant RNeasy mini kit (Qiagen) and the mRNA was isolated using a custom protocol (see Methods 2 and 3 in Supplementary Material)."

From this it appears that they were using one library kit, but they were doing library preparation in several different ways.
this library kit is probably the ISR library kit that was inferred in many of the sample.

Is it possible to use the same library kit but change the method to change the library type?

• 0 views • link

written 3.7 years ago by kuwasenai_vegetable •

From this it appears that they were using one library kit, but they were doing library preparation in several different ways.

Which is rather odd. One prep is total RNA where as the others are mRNAseq. This seems pretty unusual and should not be directly comparable.

• 0 views • link

written 3.7 years ago by GenoMax

Thank you for pointing this out !

I checked and it was not mentioned in the paper, but looking at the GEO database, all samples were total RNA preparations.This implies that the same process performed in all samples.
Then the sample that was inferred to be IU is suspected to have problems with the sample data, but the paper said:
"All raw sequencing read libraries were checked with FastQC to ensure their sequencing quality (62), and all samples passed the quality requirements."
In the end, I'm back to the first question.

• 0 views • link

written 3.7 years ago by kuwasenai_vegetable •

It is possible that one of the samples is bad. You can't control for this when using public data.

• 0 views • link

written 3.7 years ago by GenoMax

0 answers

No answers yet.