How to interpret Corona Lite error file

[begsch]

Hi everybody,

I am new to Corona Lite. The output contained in the error file '.unique.tag_length.missmatch_number.errors.csv' is not clear to me. The documentation of Corona Lite tells only that this file contains the "Error profile for the tag generated from unique matching result.". Therefore I am addressing myself to the Corona-Gurus to get to know more about it. Thank you very much in advance!

Here an example error output (as reference only 1 single sequence was used):
0_CA,2945
0_CG,666
0_AT,3762
1_03,549
1_30,1903
1_31,930
1_32,269
2_01,2478
2_02,1103
2_03,3919
...
42_30,55184
42_31,63875
42_32,48066
43_01,45960
43_02,39217


Best regards

[westerman]

I am not a corona-lite expert but do use the program a lot for our SOLiD runs.

The file simply contains the type and number of errors found at each position of all of the reads. In your above example, for position #1 there were four types of errors found: 01, 30, 31, and 32 colorspace transitions with the corresponding count of reads with that type of error (e.g., 549, etc.)

For your example the number of errors for position #1 is rather low while for position #42 and #43 is getting higher. But being at the end of a read this is expected.

One expects errors. There will be sequencing errors. There will be adjacent errors from SNPs. Because of when the above file is generated within the Corona pipeline I believe but am not sure that the file will show all errors -- sequencing and SNP. Depending on how far away your reference sequence is evolutionarily from your reads then you may get quite a few errors (my most recent SOLiD analysis has a lot more errors in many different types of transitions than you indicate but I expect that because of my reference).

You can use the file in at least a couple of ways to raise red flags about the run. First, if a position has a large number of errors compared to positions around it then that would indicate that the particular position had some sort of sequencer errors. Second if, for a particular position there is an large imbalance in the transition that that would indicate a possible problem. Using the above example I suspect that most of your reads mapped to '33' in position one with some of them having the erroneous 30, 31, 32 and 03 colorspace transitions. The count of those four are fairly well balanced. But say that the '32' transition was 50 times the count of the other transitions. What would that say about the mapping, the reference, and the reads? If there was a large '2X' transition in position 2 then I would say 'strong SNP' ... but there may be other problems indicated by this.

Anyway as I said, I am not an expert so if someone else comes up with a better explanation then use that.

[begsch]

Hi westerman,

Thanks a lot for your description. The output is now much clearer to me. Meanwhile I did some more test runs using other short read sources and reference genomes. I can also state transitions with very high error counts -as you have mentioned- compared to other transitions for the same read position. I have read about SHRiMP and its capacities to detect SNPs better than the mapper of Corona Lite - at least for the Ciona savignyi example of the SHRiMP paper. Have you tested SHRiMP so far? If yes, what do you personally prefer as color-space mapper?

Thanks again

[nilshomer]

Hi westerman,

Thanks a lot for your description. The output is now much clearer to me. Meanwhile I did some more test runs using other short read sources and reference genomes. I can also state transitions with very high error counts -as you have mentioned- compared to other transitions for the same read position. I have read about SHRiMP and its capacities to detect SNPs better than the mapper of Corona Lite - at least for the Ciona savignyi example of the SHRiMP paper. Have you tested SHRiMP so far? If yes, what do you personally prefer as color-space mapper?

Thanks again

You should also take a look at BWA, MAQ, and my own aligner BFAST (biased) for color space data. The advantages of SHRiMP and BFAST is they find indels within the reads using dynamic programming, whereas for example corona-lite requires additional tools to find the indels.