This question might be silly but I need some clarification.

I'm slightly confused about if there are multiple PAM and BLOSUM matrices? Is there a different matrix for each sequence?

My understanding is that PAM and BLOSUM are substitution matrices that are used to demonstrate how similar a sequence is to other sequences by seeing how many iterations of the matrix multiplication are needed before being able to achieve a "similar primary composition". Is this correct?

As I've understood it this means that if I want to compare 2 sequences together I can just this iterative calculation until the relationship between my target sequence and my initial sequence is maximized and then see if that similarity is enough that I can say they are closely related.

This is all assuming there is one standard PAM250 and BLOSUM62 matrix which I'm assuming are calculated from information derived from all information collected on protein sequences.

So when using PSIBLAST to search for similar sequences using BLOSUM matrices does it calculate similarity scores using BLOSUM matrices and then if this score is above a certain threshold it returns that sequence. If you do this for multiple iterations then what is the next input that it uses?


EDIT: I think some of the reason for my misunderstanding is that I'm not sure why there would be one single matrix which can be used to determine the "resilience" of a particular amino acid in general. Isn't the "usefulness" or that amino acid dependent on the function of the protein? In that case shouldn't there be different substitution matrices for different proteins? Is this why PAM matrices have difficulty and is this semi-solved by using BLOSUM matrices which use blocks and so can capture when certain "blocks" of amino acids correspond to particular functions and are therefore seen together a lot and are resilient to change in that way? It still seems a bit strange to lump them together in this case.

  • $\begingroup$ There are multiple matrices. The matrices scores various mutations slightly different and the use of them depends on how big the evolutionary distance is expected to be among the sequences. (en.wikipedia.org/wiki/Point_accepted_mutation)(https://… $\endgroup$ – Jeppe Nielsen Apr 25 '17 at 11:58
  • 1
    $\begingroup$ @JeppeNielsen There are multiple matrices as the matrix is generated based on the sequence that I'm querying? Or there are a few standard matrices that are used depending on the type of query? $\endgroup$ – tryingtolearn Apr 25 '17 at 12:07
  • 1
    $\begingroup$ Possible duplicate of Scoring matrices (BLOSUM & PAM) in BLAST and other sequence-comparison programs $\endgroup$ – David Apr 25 '17 at 14:42
  • $\begingroup$ Possible duplicate of biology.stackexchange.com/questions/49036/…. Please search for other questions on a topic on SE Biology. BLOSUM brings up several hits. $\endgroup$ – David Apr 25 '17 at 14:44
  • 1
    $\begingroup$ The Wikipedia entries describe the data used for these matrices. As already commented, many pre-existing Protein (not DNA as in your question) sequences are needed and used to get a meaningful matrix. No, it isn't perfect, but your intuition is wrong — it works well enough. No, you can't use a matrix from two sequences, and especially two you are trying to align with a program that uses a matrix. Yes, you can modify standard matrices to get position-specific matrices as you perform multiple alignments of many sequences you already know are similar (see docs for Clustal). $\endgroup$ – David Apr 25 '17 at 15:44

Your Answer

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

Browse other questions tagged or ask your own question.