Illumina-Tag sequencing, filtering homopolymers/entropy based filtering

[tonybert]

Hi all, I have a general question to anyone that has been involved in tag-sequencing of functional genes in environmental samples. I use a data processing pipeline based on Usearch (Rob Edgar's methods) and QIIME to analyze functional genes related to the nitrogen cycle in soils and marine environments . I am using usearch, specifically the fastq_stats option, to assess quality values of my sequences over read-length. After visusal inspection of read-length vs. Q value, I am then trimming the sequences to a fixed position, and using QIIME to filter reads using split_libraries_fastq.py with these quality options (-q 19 -r 3 -p 0.75 -n 0). I am using QIIME rather than usearch exclusively because I like to then split the cleaned file with split_fasta_on_sample_ids.py (i have lots of samples, some from different environments, that I would like to analyze independently). Anyway, I have found that for some target genes of lower abundance organisms, there are often a substantial proportion of sequences still in the cleaned files with homo-polymer runs of AAAAAAA, or something similar to this. After doing some investigation, I found that there are ways to eliminate these sequenes using DUST filters, such as those found in the PrinSeq package. However, I am wondering if anyone knows of methods, scripts, or filter options within QIIME or usearch that will remove these sequences. The reason I am concerned about these files is obvious, I hope, and a substantial amount of these reads ends up contributing to my OTU files, yet display no hits to reference gene databases of my targets or even GenBank!

So, overall, does anyone know if entropy based filters exist in QIIME or usearch for elimination of low-complexity sequences.

Hope my question makes sense!

Thanks,

-Tony