Scythe and all supporting documentation Copyright (c) Vince Buffalo, 2011-2014

Contact: Vince Buffalo vsbuffaloAAAAA@gmail.com (with the poly-A tail removed)

If you wish to report a bug, please open an issue on Github (http://github.com/vsbuffalo/scythe/issues) so that it can be tracked. You can contact me as well, but please open an issue first.

Scythe uses a Naive Bayesian approach to classify contaminant substrings in sequence reads. It considers quality information, which can make it robust in picking out 3'-end adapters, which often include poor quality bases.

Most next generation sequencing reads have deteriorating quality towards the 3'-end. It's common for a quality-based trimmer to be employed before mapping, assemblies, and analysis to remove these poor quality bases. However, quality-based trimming could remove bases that are helpful in identifying (and removing) 3'-end adapter contaminants. Thus, it is recommended you run Scythe before quality-based trimming, as part of a read quality control pipeline.

The Bayesian approach Scythe uses compares two likelihood models: the probability of seeing the matches in a sequence given contamination, and not given contamination. Given that the read is contaminated, the probability of seeing a certain number of matches and mismatches is a function of the quality of the sequence. Given the read is not contaminated (and is thus assumed to be random sequence), the probability of seeing a certain number of matches and mismatches is chance. The posterior is calculated across both these likelihood models, and the class (contaminated or not contaminated) with the maximum posterior probability is the class selected.

Scythe can be compiled using GCC or Clang; compilation during development used the latter. Scythe relies on Heng Li's kseq.h, which is bundled with the source.

Scythe requires Zlib, which can be obtained at http://www.zlib.net/.

To build Scythe, enter:

make all

Then, copy or move "scythe" to a directory in your $PATH.

Scythe can be run minimally with:

scythe -a adapter_file.fasta -o trimmed_sequences.fasta sequences.fastq

By default, the prior contamination rate is 0.05. This can be changed (and one is encouraged to do so!) with:

scythe -a adapter_file.fasta -p 0.1 -o trimmed_sequences.fastq sequences.fastq

If you'd like to use standard out, it is recommended you use the --quiet option:

scythe -a adapter_file.fasta --quiet sequences.fastq > trimmed_sequences.fastq

Also, more detailed output about matches can be obtained with:

scythe -a adapter_file.fasta -o trimmed_sequences.fasta -m matches.txt sequences.fastq

By default, Illumina's quality scheme (pipeline > 1.3) is used. Sanger or Solexa (pipeline < 1.3) qualities can be specified with -q:

scythe -a adapter_file.fasta -q solexa -o trimmed_sequences.fasta sequences.fastq

Lastly, one can specify the minimum match length argument with -n <integer> and the minimum length of sequence (discarded less than or equal to this parameter) to keep after trimming with -M <integer>:

scythe -a adapter_file.fasta -n 0 -M 10 -o trimmed_sequences.fasta sequences.fastq

The default is 5. If this pre-processing is upstream of assembly on a very contaminated lane, decreasing this parameter could lead to very liberal trimming, i.e. of only a few bases.

Note that the two provided adapter sequence files contain non-FASTA characters to denote the locations of barcode sequences, which always appear in TruSeq adapters, and may or may not appear in forward and/or reverse reads using the original Solexa/Illumina adapter sequences, depending on library preparation. You'll need to modify the adapter sequence files in order to use them.

In the case of the original Solexa/Illumina adapter sequences, we've seen barcodes "upstream" of forward reads (in which case the reverse complement of the barcode will appear before the adapter sequence at the 3'-end of reverse reads - replacing the [NNNNNN]). We've also seen barcodes upstream of reverse reads (in which case the reverse complement of the barcode will appear before the adapter sequence at the 3'-end of forward reads - replacing the [MMMMMM]). Your definition of the barcode may be someone else's reverse-complemented barcode, and the barcode may or may not be 6 bases.

In the case of TruSeq adapter sequences, there will always be a 6 bp barcode in place of the [NNNNNN] in sequence contaminating forward reads (if the fragment is short enough, of course). This barcode sequence should match the barcode included in the reads' FASTQ headers.

Scythe only checks for 3'-end contaminants. As of commit 7f49366, the algorithm has changd, and Scythe now matches 3' contaminants up to the 5'-end. Still, do not use Scythe for 5'-trimming as (1) this is a trivial problem for most Illumina sequences, as the quality is high in this region, and (2) Scythe does not allow for adapters overlap the 5'-end (though this may come in the future).

For reads with contamination in any position, the program TagDust (http://genome.gsc.riken.jp/osc/english/dataresource/) is recommended. Scythe has the advantages of allowing fuzzier matching and being base quality-aware, while TagDust has the advantages of very fast matching (but allowing few mismatches, and not considering quality) and FDR. Note that TagDust removes contaminated reads entirely, while Scythe trims off contaminating sequence, leaving valuable uncontaminated read sequence!

A possible pipeline would run FASTQ reads through Scythe, then TagDust, then a quality-based trimmer, and finally through a read quality statistics program such as qrqc (http://bioconductor.org/packages/devel/bioc/html/qrqc.html) or FASTqc (http://www.bioinformatics.bbsrc.ac.uk/projects/fastqc/).

Scythe does work with paired-end data. Each file must be run separately, but Scythe will not remove reads entirely leaving mismatched pairs.

In some cases, barcodes are ligated to both the 3'-end and 5'-end of reads. 5'-end removal is trivial since base calling is near-perfect there, but 3'-end removal can be trickier. Some users have created Scythe adapter files that contain all possible barcodes concatenated with possible adapters, so that both can be recognized and removed. This has worked well and is recommended for cases when 3'-end quality deteriorates and prevents barcode removal. Newer Illumina chemistry has the barcode separated from the fragment, so that it appears as an entirely separate read and is used to demultiplex sample reads by Illumina's CASAVA pipeline.

No. Embracing the Unix tool philosophy that tools should do one thing very well, Scythe just removes 3'-end contaminants where there could be multiple base mismatches due to poor base quality. N-mismatch algorithms (such as TagDust) don't consider base qualities. Scythe will allow more mismatches in an alignment if the mismatched bases are of low quality.

Scythe only checks as far in as the entire adapter contaminant's length. However, some investigation has shown that Illumina pipelines sometimes produce reads longer than the read length + adapter length. The extra bases have always been observed to be A's. Some testing has shown this can be addressed by appending A's to the adapters in the adapters file. Since Scythe begins by checking for contamination from the 5'-end of the adapter, this won't affect the normal adapter contaminant cases.

For each adapter in the file, the contaminants removed by position are returned via standard error. For example:

Adapter 1 'fake adapter' contamination occurences:
[10, 2, 4, 5, 6]

indicates that "fake adapter" is 5 bases long (the length of the array returned), and that there were 10 contaminants found of first base (-n was set to 0 then), 2 of the first two bases, 4 contaminants of the first 3 bases, 5 of the first 4 bases, etc.

No, as these have no quality information.

See the section below.

As far as I know, Scythe is the only program that employs a Bayesian model that allows prior contaminant estimates to be used. This prior is a more realistic approach than setting a fixed number of mismatches because we can visually estimate it with the Unix tool less.

Scythe also looks at base-level qualities, not just a fixed level of mismatches. A fixed number of mismatches is a bad approach with data our group (the UC Davis Bioinformatics Core) has seen, as a small bad quality run can quickly exhaust even a high numbers of fixed mismatches and lead to higher false negatives.

Just link to this Github repository, until I finish the manuscript (sorry).

Scythe is free software and is proved without a warranty. However, I am proud of this software and I will do my best to provide updates, bug fixes, and additional documentation as needed. Please report all bugs and issues to Github's issue tracker (http://github.com/vsbuffalo/scythe/issues). If you want to email me, do so in addition to an issue request.

If you have a suggestion or comment on Scythe's methods, you can email me directly.

I am currently writing a paper on Scythe's methods. In my preliminary testing, Scythe has fewer false positives and false negatives than its competitors.