Accuracy estimated by quiver

[macb]

Hi !

Does the QV reported by quiver concerns the genome accuracy before or after polishing ? Then, do lower cases bases are counted in the QV calculation ?

Best

[rhall]

The QV isn't "genome accuracy" it is the probability of a base being correct in the consensus given the alignment. Lowercase bases are none called bases and are reported as the reference sequence therefore do not have a QV.

[macb]

So the consensus accuracy is equal to the mean of all called bases QV ?

[rhall]

No the QV of the consensus base is determine by the quiver algorithm and is a function of the quality of the alignment at that point.

[macb]

Ok Sorry for this confusion. I have seen somewhere that it is recommended to iterate quiver until QV > 50. So I guess that it should concern all the bases or may be the mean bases QV ?

[gconcepcion]

Ok Sorry for this confusion. I have seen somewhere that it is recommended to iterate quiver until QV > 50. So I guess that it should concern all the bases or may be the mean bases QV ?

You're talking about different QVs. The QVs that are produced by quiver are per-base QVs.

When someone says to iterate quiver until the QV is >50, they are referring to the QV estimated by aligning an entire sequence to a reference, and calculating an error rate.

e.g. a QV of 50 would imply that there is 1 error in a 100,000 Bp window between the sequence in question and the reference.

btw> I don't necessarily agree that iterating quiver until QV is >50 is the correct strategy. How can you be sure of the accuracy of the reference?

[gconcepcion]

Just to clarify, internally we refer to this as a "Concordance QV", not an "Accuracy QV". You are calculating how well a particular sequence "concords" with a particular reference sequence. Using the term "Accuracy" implies that the reference is the absolute truth, when in reality it's just one version of an assembly that was chosen to map against.