Analyzing mutational bias in single gene

[simon_seq]

Hello,
I've got a small bioinformatics problem that I'm sure is easy for many of you to solve.

For my project I create mutant libraries of a gene. For instance within 300 nucleotides I have 5 codons [5x3 bases], which are randomized to any other codon. This gene library I then take and expose to a treatment condition to identify mutants, which are enriched under the given condition.
To analyze these samples I perform Illumina MiSeq 2x150 sequencing. In order to identify the bias within the sites, which I know to be randomized, I match all reads against a pattern using regular expressions. This analysis however does not tell me if additional mutations have occurred during the experiment apart from the randomized codons. In order to get this information I map the sequencing data against a 300bp reference sequence using BowTie and then visualize the resulting alignment in the Broad's IGV. Now the problems I have are the following:
1) BowTie only find alignments for 5% of the entire dataset (pattern matching returns 40% exact matches).
2) BowTie breaks up the 150bp reads and separately aligns fragments of ~70bp (at times multiple for one read if -a option is used)
3) The read breaks appear to occur specifically at sites, which are randomized.

Is there any way to tell BowTie to leave reads intact and accept three consecutive mutations?

Thanks a lot for your help!
Simon

[swbarnes2]

I'm not sure Bowtie is the right tool for the job. You are after all aligning to a very short sequence, not a huge genome.

I think you should compress your fastq down to unique sequences (counting how many times each sequence is in your fastq) and then use a more sensitive aligner to figure out what each of those unique sequences means.

[simon_seq]

Hi swbarnes2,

Thanks for your reply. It's probably reasonable to suggest a more sensitive aligner. The reason I've chosen bowtie is its performance on large datasets, not requiring any compressing. Would you by any chance have a suggestion for a more ideal aligner in this case?
Regarding the compressing, are you suggesting using FASTQ/A Collapser? (http://hannonlab.cshl.edu/fastx_toolkit/) Would the output be compatible with most alignment tools?

Thanks and best wishes
Simon