RAW and Processed data

[khushal]

I got (illumina next seq 500) 2x150 bp paired end sequenced mouse transcriptome data.
The vendor gave me a QC report of both Raw and processed data.
In RAW data the max. and min. read length is 151 bp(both) whereas in peocessed data (processed by vendor through NGS QC tool kit) the min and max. read length is 50 and 151 bp respectively.How ? my confusion is that raw data should have all the possible read lengths from 50 to 151 bp ?
Second, i did Fast QC of both, RAW and processed data. Against sequence length distribution window of FastQC,I found normal result with RAW data whereas warning was issued with processed data.

Should i go for down steam processing with RAW or processed data ?

[arthurmelo]

Normally all reads has the same length on raw sequenced data. It explain the equal values of min and max read length. Generally the raw dataset is parsed by some bioinformatic tool like trimmomatic in order to eliminate nucleotides with low quality of base call using Phred scores. This Phred treshold could have different values, but Q30 (accepted one error by thousand of bp) is recommended to call variants and DGE analysis, for example. Its also common delete all reads based in a minimum length required and I supoosed this minimum length is 50bp in your case. So, after this initial filter the processed or parsed dataset has the quality data you should be used to forward analysis.

[dpryan]

The raw data is always a single read length. The processing involves adapter and quality trimming, which is why there's then a range of lengths. You should use the processed results for most common downstream applications.

[dpryan]

Q30 (accepted one error by thousand of bp) is recommended to call variants and DGE analysis, for example.

FYI, this is no longer recommended. For DGE, trimming above Q5 is rarely beneficial. For variant calling the callers themselves take the phred score into account, so there's never a need to trim so stringently.