Butter: High-precision genomic alignment of small RNA-seq data

Abstract

Eukaryotes produce large numbers of small non-coding RNAs that act as specificity determinants for various gene-regulatory complexes. These include microRNAs (miRNAs), endogenous short interfering RNAs (siRNAs), and Piwi-associated RNAs (piRNAs). These RNAs can be discovered, annotated, and quantified using small RNA-seq, a variant RNA-seq method based on highly parallel sequencing. Alignment to a reference genome is a critical step in analysis of small RNA-seq data. Because of their small size (20-30 nts depending on the organism and sub-type) and tendency to originate from multi-gene families or repetitive regions, reads that align equally well to more than one genomic location are very common. Typical methods to deal with multi-mapped small RNA-seq reads sacrifice either precision or sensitivity. The tool ‘butter’ balances precision and sensitivity by placing multi-mapped reads using an iterative approach, where the decision between possible locations is dictated by the local densities of more confidently aligned reads. Butter displays superior performance relative to other small RNA-seq aligners. Treatment of multi-mapped small RNA-seq reads has substantial impacts on downstream analyses, including quantification of MIRNA paralogs, and discovery of endogenous siRNA loci. Butter is freely available under a GNU general public license.

Evaluation of microRNA alignment techniques

“…Bowtie1 with default settings demonstrated poor accuracy in most tests and provided the third lowest accuracy overall. A recent report has highlighted considerable strand-bias in Bowtie1 when using the default settings and cautioned that this is likely to have negatively impacted previously published analyses (Axtell 2014). In contrast, Bowtie1 using the "best strata" setting was highly accurate in our tests (including multimapped reads), and is used in microRNA prediction pipelines such as miRDeep2 and microRNA target prediction (StarBase) (Friedländer et al 2008;Hackenberg et al 2011).…”

Comparison of small RNA profiles in Nicotiana benthamiana and Solanum lycopersicum infected by polygonum ringspot tospovirus reveals host-specific responses to viral infection

“…In these latter interactions, activity of the silencing-suppressor proteins P19 and P38, which, respectively, sequesters 21 nt duplex siRNAs (Vargason et al, 2003), and blocks AtDCL4 activity (Azevedo et al, 2010), could determine the observed high accumulation of 22 nt vsRNAs. In this context, a specific activity and/or differential behavior of PolRSV NSs in the two hosts could have a role in determining the different vsRNA profile.…”

“…Clean reads were used for mapping small RNA sequences against the viral (GenBank accession numbers: KJ575619 for the L segment, HQ830188 for the M segment, HQ830187 for the S segment, Margaria et al, 2014), and host genomes (N. benthamiana genome version 0.5, and tomato genome version 2.50) (Naim et al, 2012;Tomato Genome Consortium, 2012), using butter version 0.3.3 (Axtell, 2014), allowing zero mismatches. Read size, distribution along the genomic segments and along single ORFs, polarity, 5 -nt enrichment and hotspot analyses were performed using samtools version 1.0 (Li et al, 2009) and in-house perl scripts, and imaged using Microsoft Excel ® v. 10, exactly as previously described (Margaria et al, 2015a).…”

Small RNA profiles of wild-type and silencing suppressor-deficient tomato spotted wilt virus infected Nicotiana benthamiana

“…The following Genbank accessions were used for isolate p202/3WT (KJ575619 for the L segment, HQ830188 for the M segment, HQ830187 for the S segment), and p202/3RB (KJ575619 for the L segment, HQ830185 for the M segment, HQ830186 for the S segment). Alignment against plant host genome was also performed using butter (version 0.2.5) (Axtell, 2014), using N. benthamiana genome version 0.5 (Naim et al, 2012) allowing zero mismatches. Reads size and distribution along the viral genome were determined using samtools (version 1.0) (Li et al, 2009) and custom perl scripts.…”

“…Adapter sequences (5 adapter sequence: TTCAGAGTTCTACAGTCCGACGATC, 3 adapter sequence: TCGTATGCCGTCTTCTGCTTG) were removed using a custom BGI script, allowing four mismatches. Trimmed reads were aligned against the viral genomic sequences deposited in GenBank (see below) using butter (version 0.2.5) (Axtell, 2014), allowing zero mismatches. The sequences of the three genomic segments of isolate p105 (Genbank accessions KJ575620 for the L segment, KJ575620 for the M segment, DQ376178 for the S segment), were used for the analysis of the WT and the derived mutant strain p105RBMar.…”