Tool to deconvolve chromatogram?

[pag]

I posted the following at another forum and have not yet had a response. Is this what mpileup and/or GATK are for? Or do those analyze aligned data for indels and SNPs, rather than look at the raw data?
====
I don't know if I'm using the proper terminology, but I have ab1 (sanger) sequencing chromatograms and I was wondering if there was any software out there that you were aware of to deconvolve two overlapped chimeric reads (e.g. overlaid indel-type mutations) and output two+ sequences as (fasta) output.

For example, if I have a read that is unambiguously ACTGGCGA but then I have 60% population where I have an A and 40% where that A is deleted, followed by GCGTGA, phred will likely give me ACTGGCGAAGCGTGA, but is there a way for a basecaller to give me ACTGGCGAGCGTGA as a secondary call? Or, if I have 50/50 and I know that one version is ACTGGCGAAGCGTGA, supplying that as a comparison file for subtraction, leaving me with a residual signal of either the full ACTGGCGAGCGTGA or of GCGTGAx?

Obviously I would have to be able to set a threshhold where I consider the result "noise" - either a fixed value of signal strength or e.g. 5% of the main peak strength in order to filter out illegitimate base calls of non-chimeric sequence.

As a separate, but related, issue, is there a way to get phred (or any other program) to "filter" noise spikes in chromatograms? For example, sometimes I see spikes in pyrimidine signal strength that is way out of proportion to legitimate regions of call (the peak height using an ab1 viewer like bioedit or consed - I'm not sure if linear or log scale - is well over double of any nearby base or even any other place in the file. These typically have "width" of about 5 base calls)

[krobison]

There are numerous programs in the Sanger world to solve this; Mutation Surveyor is one I have used but there is also polyphred. Most commercial packages for Sanger data should attempt to do this.

To answer your original question, GATK, mpileup and the like don't work directly with chromatograms, though if you convert them to FASTQ format they can -- but then you have lost the information in this case you are looking for.

This underscores a key difference between Sanger and the newer technologies -- Sanger can generate a signal from a mixed population which you can attempt to deconvolve. The newer technologies are all looking at individual molecules (sometimes by amplifying them into a population to do so), and so this form of deconvolution just isn't relevant.

[pag]

Thanks!
I'll try out polyphred, given that I have the package that it interfaces with.