Error rate in BBMAP

[mido1951]

hello,
I want to know the equation error rate in BBMAP?
how you calcuer that rate?
Thank you for your reply.
@Brain?

[mido1951]

any response for this?

[HESmith]

What do you mean by "equation error rate"? Are you asking what fraction of reads are aligned to the incorrect loci? Or how errors affect the alignment accuracy? Or something else?

[mido1951]

no, I want to know how you get the error rate?
what is the distance to get the error rate?
formally, how to express the error rate in BBMAP?

Error Rate:              23.1932%          281318         [B]3.9989%[/B][I](this)[/I]           24419957

[HESmith]

Again, to what error rate are you referring? There are multiple errors that can be measured, such as the two examples I provided.

What command(s) did you use to obtain the line of code in your previous post?

[Brian Bushnell]

Hmmm, I think mido is referring to the stderr output of BBMap after it finishes running. Those columns are:

(name), (% of reads with any errors), (number of reads with any errors), (% of bases with any errors), (number of bases with any errors).

Please note that the value can be a bit misleading if a lot of reads are mapped with long deletions. For Illumina reads, it's better to look at the substitution rate unless you reduce BBMap's default "maxindel" flag from the default 16000 down to a much lower value of perhaps 100.

The way it is calculated is based on the number of total alignment operations and number of matching alignment operations. Internally, when BBMap aligns a read to a reference, it supports 5 operations:

M: Match
S: Substitution (a base in the read differs from the reference)
I: Insertion (a base in the read not present in the reference)
D: Deletion (a base in the reference not present in the read)
N: No-call (undefined in the read or the reference)

These roughly correspond to cigar strings, but cigar strings do not have an equivalent of the "N" symbol, and they do have a lot of strange, poorly-defined symbols, rendering them not very useful in computation.

Simply, the sub rate is calculated as S/(M+S+I+D+N). The del rate is D/(M+S+I+D+N), and so forth. The error rate is (S+I+D+N)/(M+S+I+D+N).

For example, this match string:

mmmmSmmmmImmmmDDDDDDmmmmN

...has 16 matches, 1 sub, 1 insertion, 6 deletions, and 1 N, for 25 total operations. The cigar string would be 4=X4=I4=6D4=X, or something like that (the specification is not fully defined). In this case the sub rate would be 4%, ins rate 4%, del rate 24%, and N rate 4%, giving an error rate of (1+1+6+1)/(16+1+1+6+1)=36%.

[mido1951]

(name), (% of reads with any errors), (number of reads with any errors), (% of bases with any errors), (number of bases with any errors).

thankyou Brian for your Explanations.
I made a mapping with BBmap and i saw the error rate (% of bases with any errors) because you have told me the other day that we must take account of this error rate.
I am looking how you calculate that error rate and what is the equation because I need to put it in my research.
thank you

[GenoMax]

Since you mostly appear to be working with MinION data is this question for data aligned with mapPacBio.sh?

If you mapped your MinION data with regular bbmap.sh then the errors could be different.

[mido1951]

Since you mostly appear to be working with MinION data is this question for data aligned with mapPacBio.sh?

If you mapped your MinION data with regular bbmap.sh then the errors could be different.

I did not use mapPacBio.sh because i have Minion reads.
But I have corrected reads and I made a mapping these reads to the reference genome and I see the error rate.
but I have to express the error rate and I have to put it in my research because I can not put an error rate without turning the equation of error rate.

[GenoMax]

You should have used mapPacBio.sh with raw MinION reads but that is not the main point here.

@Brian: Did provide an explanation of how the error rate is calculated above (with the equation). Numbers we see in the final output must be an average across all mapped reads.

[mido1951]

the global error rate equation is:
The error rate is (S+I+D+N)/(M+S+I+D+N) ??

[GenoMax]

Yes I think so. Per @Brian

The way it is calculated is based on the number of total alignment operations and number of matching alignment operations.

[mido1951]

what's the distance used in every M, D, I,....??

[GenoMax]

@Brian will have to chime in with a final word but I think the rate is an average across all alignment operations as he indicated above. Each read will have its own M,S,I,D,N values from its CIGAR strings.

[mido1951]

Is that the distance M=N=I=D=S=1??
any response Brian?