Converting ABI colorspace qualities into base qualities

[szilva]

Thank you indeed for the answers - yes, these qualities are pretty poor, but there are some (few dozens of millions :S ) other reads with better qualities to process.

The reads are aligned, I have output files generated by SHRiMP that has an other BLAST-like format also where the reads are locally aligned. But I really want to understand this quality issue before converting results to SAM and generating a consensus. It is much clearer now, probably I will validate my findings with BFAST.

[maubp]

See also this thread, although it does not discuss the qualities:

http://seqanswers.com/forums/showthread.php?t=776

[nilshomer]

Hi, my simple problem is that considering I have a CS read with qualities like:

>1_2_3_F3
T32123202112210132121031230312312212231012221122201

>1_2_3_F3
2 2 2 2 2 5 7 2 2 4 8 4 2 5 2 2 3 2 2 2 4 3 2 2 2 2 11 2 2 3 2 4 2 2 2 8 12 2 4 2 8 5 2 2 2 10 3 2 2 2

How will I able to convert this correctly into a format (ie SAM) ? I can easily assign nucleotides to the T312... sequence but I could not really figure out how to map the color qualities to nucleotide ones.

Sorry if my question was already answered but after and hour searching I did not manage to find a proper answer - so any related docs are appreciated.

Cheers:
Szilva

Here's what BFAST and other alignment programs do:

http://sourceforge.net/apps/mediawiki/bfast/index.php?title=Mapping_Quality

If you are converting them to base space without alignment, you wont know where (if any) sequencing errors occurred. Therefore, all bases after a single sequencing error will result in an incorrect alignment. That is why it is so important to align the data.

If you want to create a SAM without converting to base space, you can store the color sequence and color qualities in the "CS" and "CQ" optional tags, while leaving the "SEQ" and "QUAL" empty (or "*").

Really, you need to tell us what is your reason to convert colors into bases.

[lvaruzza]

Use Corona Lite

You can use the script encodeFasta.py from Corona Lite package in this way:

encodeFasta.py -d file.csfasta > file.fasta

Corona is available in:

http://solidsoftwaretools.com/gf/project/corona/

Usually is not recommended to convert the reads directly to base space, it's better to map them or perform a denovo assembly. If you use bowtie, for example, it will generate a SAM file with the sequence in base space.

You don't need to change the QUAL file, it's already in phred scale.

[maubp]

OK, fine - it would have be clearer as two [ code ] blocks, that's all.

The naive answer would be to re-use the QUAL file as is. However, as I understand things, those qualities really are for color calls (transitions between bases), so that would be rather misleading. You are stuck with the fact that any color error means in sequence space all the subsequence bases will be wrong. One might therefore argue that on conversion to sequence space the quality scores should decline to reflect this cumulative probability of error.

Why do you want to try and convert this into nucleotide space quality scores anyway?

Note that if they are on the PHRED scale and this example is representative, all the quality scores are very poor (maximum 12), so this is neither here nor there. Hopefully you have some better reads.

[szilva]

Is the quoted example entries from two files? i.e. A color space FASTA file (csFASTA) and a matching color space QUAL file?

Sure, as ABI SOLiD reads are usually stored.

[maubp]

Is the quoted example entries from two files? i.e. A color space FASTA file (csFASTA) and a matching color space QUAL file?

[szilva]

Hi, my simple problem is that considering I have a CS read with qualities like:

>1_2_3_F3
T32123202112210132121031230312312212231012221122201

>1_2_3_F3
2 2 2 2 2 5 7 2 2 4 8 4 2 5 2 2 3 2 2 2 4 3 2 2 2 2 11 2 2 3 2 4 2 2 2 8 12 2 4 2 8 5 2 2 2 10 3 2 2 2

How will I able to convert this correctly into a format (ie SAM) ? I can easily assign nucleotides to the T312... sequence but I could not really figure out how to map the color qualities to nucleotide ones.

Sorry if my question was already answered but after and hour searching I did not manage to find a proper answer - so any related docs are appreciated.

Cheers:
Szilva