Quality of single sequencing read

[elfuser]

What is the best way to represent the quality of a single sequencing read with one value? I know the phred scores of each base, but I am looking to summarize that into one value. I thought the average could be an option or perhaps adding the scores. Is there a standard or which one would be more informative?
Thanks so much

[elfuser]

I saw in a post that a way of representing this could be by calculating the percentage of q20 or q30 bases in the read. But I think that this may not be as sensitive. Any suggestions

[TonyBrooks]

I thought the consensus is the % of bases >Q30 at the final base. You'll also need to know the read length as quality drops each cycle so any Q-score without a read-length is meaningless. You would expect a higher % for shorter reads.

[elfuser]

I am using illumina PE where I think the read length is always the same. Now I'm guessing you take the final base into account because the quality is always the same or drops from the beginning of the read to the end. Thanks

[TonyBrooks]

Correct. The quality should alway drop once passed the initial few bases where they appear to increase (an artefact of the RTA software). I think if you go to the Illumina website they give you estimated specs of each instrument described as % bases >Q30 for a few types of read variety.

[pmiguel]

What is the best way to represent the quality of a single sequencing read with one value? I know the phred scores of each base, but I am looking to summarize that into one value. I thought the average could be an option or perhaps adding the scores. Is there a standard or which one would be more informative?
Thanks so much

Well, the quality of a given base is and estimation of its probability of being a miscall.

So you could represent the probability of a read having zero, one, two or whatever errors in it or less. For example, trivially, take a 100 base read with quality values that are all 30. Chance of having zero errors would be 0.999^100 would be about 90.5%. More generally you can convert all the quality values into error probabilities and use their product to tell you what the chance of the read containing a miscall is.

--
Phillip

[mathew]

quality of single end reads

Is it quite common to have very high number of duplicates (PCR) in single end read should we also drop them or not I am taking about 40% of total reads in an RNA-seq.