Motif count number in MC tag not obvious from the alt sequence

[rollf]

Hi,

I have the following call (checked with 1.0.0 & 1.5.1):

chr9 69037261 . TAAAAAATACAAAAAAAAAAAAAAAAGAAGAAGAAGAAGAAGAA TAAAAAATACAAAAAAAAAAAAAAAAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA,TAAAAAATACAAAAAAAAAAAAAAAAAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA 0 . TRID=FXN;END=69037304;MOTIFS=A,GAA;STRUC=(A)n(GAA)n GT:AL:ALLR:SD:MC:MS:AP:AM 1/2:83,84:78-85,79-87:17,16:26_19,27_19:0(0-26)_1(26-83),0(0-27)_1(27-84):0.975904,0.97619:.,.

The corresponding repeat is

chr9	69037261	69037304	ID=FXN;MOTIFS=A,GAA;STRUC=(A)n(GAA)n

Here is the data in a more readable way:

GT :AL   :ALLR       :SD   :MC         :MS                                 :AP            :AM
1/2:83,84:78-85,79-87:17,16:26_19,27_19:0(0-26)_1(26-83),0(0-27)_1(27-84):0.975904,0.97619:.,.

ref TAAAAAATACAAAAAAAAAAAAAAAAGAAGAAGAAGAAGAAGAA
al1 TAAAAAATACAAAAAAAAAAAAAAAAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA
al2 TAAAAAATACAAAAAAAAAAAAAAAAAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA

ref TAAAAAATACAAAAAAAAAAAAAAAA  GAAGAAGAAGAAGAAGAA
al1 TAAAAAATACAAAAAAAAAAAAAAAAA GAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA
al2 TAAAAAATACAAAAAAAAAAAAAAAAAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA

Counting the occurences of GAA, I find indeed 19 in both alternative alleles (and they are 100% pure).
However, I don't understand the mentioned number of 26 (alt1) or 27 (alt2) occurrences of A.
There are 17 (alt1) or 18 (alt2) occurrences of A between TAC and the first occurrences of GAA; then there are 6+1 additional As at the beginning of the sequence.

Taking into all those As for alt1, the calculation is 6+1+17=24. This is somewhat near 26 * 0.975904 = 25.373504, but seems still wrong. Is my understanding correct here? What explains the numbers 26 and 27?

This issue might be related.

[pbsena]

Hello,

Thank you for reporting this output. I think what is happening here is that TRGT is estimating the count of As when seeing sequences that it did not expect. The structure for FXN is runs of As and runs of GAAs, so bases T, T and C get clustered into the runs of As. Under the hood TRGT uses a hidden markov model to find the pure sequence (runs of As + runs of GAAs) that explain the data. In this case the assignment of 26 is probabilistic. It is the hidden sequences that maximizes the likelihood of the observed data, assuming some of the Ts and Cs are mismatches and some are indels. In the homopolymer case specifically they don't always resemble the consensus. We may need to revisit this logic in a further iteration, I completely agree that the report may warrant some confusion.

[rollf]

Hi,

thanks for the quick feedback and the explanation. The delta between alt1 & alt2 is consistent with the reported alt sequences (i.e 26 + 1xA = 27) - this is good.

    1                        26
    |                        |
ref TAAAAAATACAAAAAAAAAAAAAAAA  GAAGAAGAAGAAGAAGAA
al1 TAAAAAATACAAAAAAAAAAAAAAAAA GAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA
al2 TAAAAAATACAAAAAAAAAAAAAAAAAA

    01                       25
    ||                       |
ref TAAAAAATACAAAAAAAAAAAAAAAA  GAAGAAGAAGAAGAAGAA
al1 TAAAAAATACAAAAAAAAAAAAAAAAA GAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAAGAA
al2 TAAAAAATACAAAAAAAAAAAAAAAAAA