For simplicity, suppose that I am penalising mismatches and gaps with -1 and adding a score +100 for each match. Does the optimum alignment of the three DNA sequences




equal to:




with the total score S = 294 + 294 - 6 = 586? In this case I sum the alignment score for each of the three pairs.

Or does it equal to:




with the score 200 - 7 = 193? In this case I am rewarding only the matches that occured across all three sequences at the same time.

Which of the two scores are the common Multiple Sequence Alignment algorithms trying to optimise for? Or if both are valid, then which optimal alignment (i.e. with respect to which definition) is preferred if realistic reward/penalise values were used?

  • 2
    $\begingroup$ Please do some reading on a topic before posting a question. This one is based on simplistic and false assumptions. A look at the WIkipedia page on MSA (en.wikipedia.org/wiki/Multiple_sequence_alignment) would make you aware of what a complex topic it is. $\endgroup$
    – David
    Apr 24 '17 at 15:52
  • 1
    $\begingroup$ Further would you please avoid or define abbreviations in the titles of questions. This helps people searching for a topic and those who may be interested in it but are not familiar with the jargon. I am altering the title to include the full name. Please do not revert it. $\endgroup$
    – David
    Apr 24 '17 at 16:47
  • 1
    $\begingroup$ A common scoring method is "sum of pairs", which is essentially the sum of scores of all pairwise alignments. The wiki page has mentioned this. You can find more discussions in Durbin's book biological sequence analysis, section 6.2. $\endgroup$
    – user172818
    Apr 25 '17 at 6:00

In your case both answers are equally valid.

The scoring matrix, for example DNAFULL (http://rosalind.info/glossary/dnafull/) usually contains various penalties for wrong alignments.

Also the introduce gap score should NEVER be set at 0 as it will cause the aligned sequences to be filled with artificial gaps!

The only thing the algoritms try to optimise towards is the score, therefore you need sane values for calculating the score and those you propose does not result in usefull alignments.

  • $\begingroup$ The scoring matrix you mention is used (to the extent that people actually use it) for pairwise alignment. At least for Clustal, a dynamic position-specific scoring system is used as successive sequences are added. Therefore the question does not relate to actual MSA — at the most is is only a weekend newspaper puzzle. $\endgroup$
    – David
    Apr 24 '17 at 16:45
  • 1
    $\begingroup$ The DNAFULL matrix can be used instead of the PAM or BLOSUM matrices for multiple alignments. The fact that it is mainly used for PAIRwise alignment doesn't exclude it from being used for MSA. $\endgroup$ Apr 24 '17 at 17:06
  • $\begingroup$ Jeppe Nielsen is right: MSA programs do use pair-wise scoring matrices. David is talking about progressive alignment, which is a heuristic to do MSA as finding the optimal solution is NP-hard. $\endgroup$
    – user172818
    Apr 25 '17 at 6:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.