[Bioc-sig-seq] SNP counting (Biostrings ?) question
Patrick Aboyoun
paboyoun at fhcrc.org
Fri Oct 31 21:44:12 CET 2008
Wolfgang,
I'm glad to hear you are finding the Biostrings package to be useful.
The pairwiseAlignment function has some helper functions that would be
useful to a SNP analysis. In particular, you may want to take a look at
the mismatchTable, mismatchSummary, summary, and consensusMatrix
functions to see if they provide the summaries you are looking for. I
also recommend you leverage the vectorized capabilities of the
pairwiseAlignment function since it will be easier to maintain your
code, provide faster execution time, and make it easier to summarize
your results. I have modified your code slightly to show you what you
can get "out of the box" from the Biostrings package without writing too
much code:
library(Biostrings)
# the substitution matrix
mat <- matrix(-3, nrow = 5, ncol = 5)
diag(mat) <- 1
mat[,5] <- 0; mat[5,] <- 0
rownames(mat) <- colnames(mat) <- DNA_ALPHABET[c(1:4,15)]
# the consensus reference sequence
ref <- DNAString("ACTTCACCAGCTCCCTGGC")
# the sample sequences; the 3rd one has no mutations, it's simply
shorter ...
samp <-
DNAStringSet(c("CTTCTCCAGCTCCCTGG","ACTTCTCCAGCTACCTGG","TTCACCAGCTCCCTG"))
# vectorized alignments using the ref as the "subject"
alignments <-
pairwiseAlignment(samp, ref, substitutionMatrix=mat,
gapOpening= -5, gapExtension= -2, scoreOnly=FALSE,
type="global")
# summaries useful for SNP analysis
summary(alignments)
mismatchSummary(alignments)
mismatchTable(alignments)
consensusMatrix(alignments, baseOnly = TRUE)
Patrick
Wolfgang Raffelsberger wrote:
> Hi Herve,
>
> thank you for your helpful answer.
> Initially I was wondering if there are already some functions/packages
> directly for my tasks available, as Bioconducter and CRAN are very
> strong resources.
>
> Since your Biostrings package provides all the basic functionalities,
> I've built a few (additional) functions (on top of Biostrings) doing the
> job.
> To be more precise, yes, I'm interested in detecting all alterations
> (mutations+ insertions + deletions). However, within the project's
> biological context mutations are the prime focus. Right now I'm rather
> in an exploratory/pilot phase for the nucleotide-alteration/SNP
> counting, so I'll probably improve & expand some of my quickly written
> code.
>
> For the sequence alignments I prefer global, since I'd rather want the
> 'best' match possible (ie the default setting with pairwiseAlignment()
> ). Concerning the choice of (custom) substitution matrices I'm not very
> decided. Right now I'm using very generic ones (like in the Biostrings
> vignette). Recently I've learned that there are several 'versions' of
> the Needleman-Wunsch algorithm (ie various improvements to the original
> one), so I'm not any more surprised of seeing different scores from
> different implementations...
> Anyway, once it's clear which sample/target (ie sequencer output) should
> be compared with which reference sequence (ie the consenus genomic
> sequence of the very region to be tested) I don't use the score any more
> and luckily my test-data contain few completely unexpected 'alien'
> sequences. For this initial step now I use matchPattern() - thank's for
> your hint- where I primarily need to test if sequences have to be read
> as inverse/complement.
> Again, after a bit more digging I found the corresponding functions and
> I'm quite happy with the performance !
>
> At the core of SNP/alterations counting I dissect the pairwise alignment
> and record what kind of changes have happened (at this level I noted
> some changes to the new BioC 2.3 implementation ...see also code below).
> Here I use (so far) strsplit() to make two vectors with individual
> nucleotides (and gaps) to extract the positions&changes from each of the
> lines of the pairwise alignment (see code below). I guess there may be
> other more elegant/faster ways to do this...
> As I'm primarily interested in 'simple mutations', so far I took a
> shortcut consisting in simply looking at the first of inserted
> nucleotide in cases of insertions (some additional lines not part of the
> code below). Maybe later I'll get back to expand this ... After all I
> end up with a simple matrix where each row corresponds to a position of
> the reference sequence and the columns contain counts of the
> events/alterations observed (where at least I'm able to tell if any
> insertions have happened at all) where I can run classical statistical
> tests.
>
> Cheers,
> Wolfgang
>
> # here the way I 'parse' pairwise alignments :
>
> library(Biostrings)
> mat <- matrix(-3, nrow = 5, ncol = 5)
> diag(mat) <- 1
> mat[,5] <- 0; mat[5,] <- 0
> rownames(mat) <- colnames(mat) <- DNA_ALPHABET[c(1:4,15)]
>
> ref <- DNAString("ACTTCACCAGCTCCCTGGC") # the consensus
> reference sequence
> samp <-
> DNAStringSet(c("CTTCTCCAGCTCCCTGG","ACTTCTCCAGCTACCTGG","TTCACCAGCTCCCTG"))
>
> # the 3rd one has no mutations, it's simply shorter ...
> # here (simplified) I already know that my sequence is oriented correctly
> alignRes <- matrix(nr=nchar(ref),nc=length(samp))
> for(i in 1:length(samp)) {
> x <- pairwiseAlignment(as.character(ref), samp[[i]],
> substitutionMatrix= mat, gapOpening= -5, gapExtension= -2,scoreOnly=F,
> type="global")
> cutsamp <- unlist(strsplit(as.character(x at subject),""))
> cutref <- unlist(strsplit(as.character(x at pattern),""))
> if(nchar(ref) != length(cutref)) {
> origRefNt <- unlist(strsplit(as.character(ref),""))
> offS <- 1
> while(origRefNt[1+offS] != cutref[1]) offS <- offS+1
> }
> out <- rep("Z",nchar(ref)) # use
> "Z" for (default) constant
> if( !identical(cutsamp,cutref)) {
> mutPos <- !(cutref== cutsamp)
> out[mutPos+offS] <- cutsamp[mutPos+offS]
> if(sum(cutref=="-")>0) out <- out[-which(cutref=="-")] # remove
> gaps in reference
> # gaps could/can be retained at this stage only to a limited degree
> by additional subsitutions
> }
> alignRes[,i] <- out
> }
>
>> sessionInfo()
> R version 2.8.0 (2008-10-20)
> i386-pc-mingw32
>
> locale:
> LC_COLLATE=French_France.1252;LC_CTYPE=French_France.1252;LC_MONETARY=French_France.1252;LC_NUMERIC=C;LC_TIME=French_France.1252
>
>
> attached base packages:
> [1] stats graphics grDevices utils datasets methods base
>
> other attached packages:
> [1] Biostrings_2.10.0 IRanges_1.0.1
>
> loaded via a namespace (and not attached):
> [1] grid_2.8.0 lattice_0.17-15 Matrix_0.999375-16
>
>
>
> Herve Pages a écrit :
>> Hi Wolfgang,
>>
>> If I understand correctly you want to compare a collection of "sample"
>> sequences against a "reference" sequence.
>>
>> There are many ways to "compare" 2 sequences and to get a score from
>> this
>> comparison: Do you want to allow indels or just substitutions (aka
>> replacements)?
>> Do you want the alignment to be global or local? Do you want the
>> score to be
>> determined by a custom substitution matrix or just to reflect the "edit
>> distance" between the 2 sequences (i.e. the minimum number of
>> replacements
>> + insertions + deletions required to transform sequence 1 into
>> sequence 2).
>>
>> If you are trying to get the edit distance between the sample sequences
>> and the reference sequence then you might want to look at this post:
>>
>>
>> https://stat.ethz.ch/pipermail/bioc-sig-sequencing/2008-July/000067.html
>>
>> Note that the 'type' argument is omitted in this post so
>> pairwiseAlignment()
>> is used to perform "global" alignments, that is, the score that is
>> returned
>> indicates the min nb of edit operations needed to transform the
>> *complete*
>> sequences in 'pattern' (your "sample" sequences) into the *complete*
>> 'subject'
>> (your "reference" sequence).
>> One nice property of "global" alignment is that it is a symetric
>> operation on
>> the 2 sequences to align. You can take advantage of this by inverting
>> the roles
>> of the "sample" and the "reference" sequences so 'samp' can be the
>> pattern
>> and 'ref' the subject. Then you can benefit from the fact that
>> pairwiseAlignment() is vectorized on the pattern.
>>
>> But then you say that you would primarily be interested in finding where
>> a given nucleotide differs from the query. This suggests that you are
>> in fact
>> looking for a much simpler form of alignment where only mismatches (i.e.
>> replacements) are allowed but no indels. Also you say that your
>> sample-sequences may start or end slightly later/earlier.
>> This means that you can use the matchPattern() function for your
>> problem.
>>
>> > m2 <- matchPattern(samp[[2]], ref, max.mismatch=3)
>> > m2
>> Views on a 19-letter DNAString subject
>> subject: ACTTCACCAGCTCCCTGGC
>> views:
>> start end width
>> [1] 1 18 18 [ACTTCACCAGCTCCCTGG]
>>
>> Then use the mismatch() function to get the position(s) of the
>> mismatche(s):
>>
>> > mismatch(samp[[2]], m2)[[1]]
>> [1] 6 13
>>
>> Cheers,
>> H.
>>
>>
>> Wolfgang Raffelsberger wrote:
>>
>>> Dear list,
>>>
>>> I would like to count the occurrence of (mostly single) nucleotide
>>> polymorphisms from nucleotide sequences.
>>> I got across the Biostrings package and pairwiseAlignment() that allows
>>> me to get closer to what I want but
>>> 1) I noticed that the score produced from pairwiseAlignment() is quite
>>> different to other implementations of the Needlaman-Wunsch alogorithm
>>> (eg in EMBOSS)
>>> 2) the score is not directly the information I 'm looking for since
>>> it's
>>> a mixture of the gaps & mismatches (and I don't see if/how one could
>>> modify that).
>>>
>>> However, I would primarily be interested in finding where a given
>>> nucleotide differs from the query (from a pairwise alignment) to some
>>> statistics on them, ie at which position I get which other element
>>> instead. Note, that my sample-sequences may start or end slightly
>>> later/earlier.
>>> Any suggestions ?
>>>
>>> Sample code might look like (of course, my real sequences are longer
>>> ...):
>>>
>>> ref <- DNAString("ACTTCACCAGCTCCCTGGC")
>>> samp <-
>>> DNAStringSet(c("CTTCTCCAGCTCCCTGG","ACTTCTCCAGCTACCTGG","TTCACCAGCTCCCTG"))
>>> # the 3rd one has no mutations, it's simply shorter ...
>>> pairwiseAlignment(ref, samp[[1]], substitutionMatrix = mat, gapOpening
>>> = -5, gapExtension = -2)
>>> alignScores <- numeric()
>>> for(i in 1:3) alignScores[i] <- pairwiseAlignment(ref, samp[[i]],
>>> substitutionMatrix = mat, gapOpening = -5, gapExtension = -2,
>>> scoreOnly=T)
>>> alignScores # the 3rd sequence without mismatches gets worst score
>>>
>>>
>>> (Based on a previous post on BioC) I just subscribed to
>>> bioc-sig-sequencing at r-project.org, but I don't know if I don't mange to
>>> search the previous mail archives (on http://search.gmane.org/) since I
>>> keep getting (general) Bioconductor messages.
>>>
>>> Thank's in advance,
>>> Wolfgang
>>>
>>>
>>> By the way, if that matters, I'm (still) running R-2.7.2
>>>
>>>> sessionInfo()
>>>>
>>> R version 2.7.2 (2008-08-25)
>>> i386-pc-mingw32
>>>
>>> locale:
>>> LC_COLLATE=French_France.1252;LC_CTYPE=French_France.1252;LC_MONETARY=French_France.1252;LC_NUMERIC=C;LC_TIME=French_France.1252
>>>
>>>
>>>
>>> attached base packages:
>>> [1] stats graphics grDevices datasets tcltk utils
>>> methods base other attached packages:
>>> [1] Biostrings_2.8.18 svSocket_0.9-5 svIO_0.9-5
>>> R2HTML_1.59 svMisc_0.9-5 svIDE_0.9-5 loaded via a
>>> namespace (and not attached):
>>> [1] tools_2.7.2
>>>
>
> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
> Wolfgang Raffelsberger, PhD
> Laboratoire de BioInformatique et Génomique Intégratives
> CNRS UMR7104, IGBMC
> 1 rue Laurent Fries, 67404 Illkirch Strasbourg, France
> Tel (+33) 388 65 3300 Fax (+33) 388 65 3276
> wolfgang.raffelsberger at igbmc.fr
>
> _______________________________________________
> Bioc-sig-sequencing mailing list
> Bioc-sig-sequencing at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing
More information about the Bioc-sig-sequencing
mailing list