MAGenTA: a Galaxy implemented tool for complete Tn-Seq analysis and data visualization

Abstract

Motivation

Transposon insertion sequencing (Tn-Seq) is a microbial systems-level tool, that can determine on a genome-wide scale and in high-throughput, whether a gene, or a specific genomic region, is important for fitness under a specific experimental condition.

Results

Here, we present MAGenTA, a suite of analysis tools which accurately calculate the growth rate for each disrupted gene in the genome to enable the discovery of: (i) new leads for gene function, (ii) non-coding RNAs; (iii) genes, pathways and ncRNAs that are involved in tolerating drugs or induce disease; (iv) higher order genome organization; and (v) host-factors that affect bacterial host susceptibility. MAGenTA is a complete Tn-Seq analysis pipeline making sensitive genome-wide fitness (i.e. growth rate) analysis available for most transposons and Tn-Seq associated approaches (e.g. TraDis, HiTS, IN-Seq) and includes fitness (growth rate) calculations, sliding window analysis, bottleneck calculations and corrections, statistics to compare experiments and strains and genome-wide fitness visualization.

Availability and implementation

MAGenTA is available at the Galaxy public ToolShed repository and all source code can be found and are freely available at https://vanopijnenlab.github.io/MAGenTA/.

Supplementary information

Supplementary data are available at Bioinformatics online.

1 Introduction

Transposon insertion sequencing (Tn-Seq) combines transposon mutagenesis on a genome-wide scale with massively parallel sequencing to determine, whether a gene, or a specific genomic region, is important for fitness under a chosen experimental condition (Gawronski et al., 2009; Goodman et al., 2009; Langridge et al., 2009; van Opijnen et al., 2009; van Opijnen and Camilli, 2013). Tn-Seq has been successfully used to identify essential genes (Chao et al., 2013; Griffin et al., 2011; Klein et al., 2012; Le Breton et al., 2015; Zomer et al., 2012), however, it is often ignored that it is possible to look further then just a binary outcome (i.e. live versus dead/very sick). Instead we have shown that the fitness of each mutant can be accurately calculated and represented as the growth rate. This approach is so sensitive that it enables identification of mutants that differ by as little as 2 min (5%) in their doubling times (van Opijnen et al., 2009; van Opijnen and Camilli, 2012). To underscore this sensitivity we have confirmed growth rates obtained from Tn-Seq for >300 in vitro and in vivo phenotypes by individual growth and 1 × 1 competitions. Consequently, besides identifying (conditionally) essential genes, this approach has enabled a range of discoveries including the identification of new leads for gene function, novel small non-coding RNAs (ncRNAs), higher order genome organization and of host-factors that affect bacterial host susceptibility (Carter et al., 2014; Jensen et al., 2016; Mann et al., 2012; van Opijnen et al., 2009, 2016; van Opijnen and Camilli, 2012). Here we present MAGenTA (Microbial Assessment by Genome-Wide Tn-Seq Analysis), a complete Tn-Seq analysis pipeline implemented in Galaxy, but also available as separate scripts. MAGenTA makes sensitive genome-wide fitness (i.e. growth rate) analysis available for most transposon and Tn-Seq associated approaches [e.g. TraDis (Langridge et al., 2009), HiTS (Gawronski et al., 2009), IN-Seq (Goodman et al., 2009)], and includes fitness (growth rate) calculations, bottleneck calculations and corrections, statistical comparisons of conditions or strains, sliding window analysis and genome-wide fitness visualization.

2 Pipeline description

MAGenTA covers the entire process of Tn-Seq data analysis from processing raw-sequence reads up to fitness calculations and data visualization, which is described in detail in Supplementary Material S1. Below we highlight major steps in the analysis that make the approach unique. The analysis is broadly split up into four parts that can be run independently. In part 1 raw data are processed and mapped to the genome, in part 2 bottlenecks are determined and corrected, and fitness is calculated for individual insertions and specified regions, in part 3 statistics are performed, data are analyzed by sliding window, and conditions or strains are statistically compared, while in part 4 data are visualized.

2.1 Read trimming, barcode splitting and mapping

Tn-Seq can be performed with different types of transposons and there are different sample preparation methods that affect what type of final product is sequenced. Some methods require specific ways in which raw reads are processed (described in Supplementary Material S1), which can be implemented here. If reads contain a multiplexing barcode they can be sorted based on this barcode, and finally reads are mapped to a reference genome.

2.2 Fitness calculations and statistics

Most Tn-Seq analyses use some form of the competition index as a proxy for fitness, however an arguably more sensitive and quantitative approach is one we developed to calculate fitness representing the growth rate (van Opijnen and Camilli, 2013). This requires sampling of two time points for sequencing and determination of the expansion or ‘contraction’ (van Opijnen and Camilli, 2012) of the population. This will generate insertion specific fitness values, and by averaging over all the insertions found within a gene or region aggregate fitness is calculated, resulting in single fitness values and standard deviations for each gene in the genome (van Opijnen et al., 2009; van Opijnen and Camilli, 2013). Importantly, this enables robust statistical analysis (see below).

2.2.1 Bottleneck calculations and corrections

A challenge that can affect Tn-Seq data, especially from in vivo experiments, is that mutants may be lost stochastically, for instance while establishing an infection. Such bottlenecks are problematic because it becomes difficult to distinguish between insertion mutants that disappear due to chance or because they are less fit. However, we have shown that by determining a set of neutral insertions (e.g. in degenerate or pseudo genes), that do not effect fitness and should therefore not disappear from the library, it is possible to estimate such bottlenecks and correct for it in the analysis (van Opijnen and Camilli, 2012).

2.3 Genome-wide sliding window analysis and fitness comparisons

Data can be further explored by performing statistics to determine which fitness changes are significant and a sliding window approach based on the methods by Zhang et al. (2012) enables the user to screen the genome for regions that have a significant fitness effect. Additionally, fitness can be compared across experiments and environmental conditions, and even between strains, with the goal to reveal strain-specific fitness effects (Jensen et al., 2016; van Opijnen et al., 2016). N.B. Two tools are specifically geared towards transposons that insert into TA sites and are marked within Supplementary Material S1.

2.4 Genome-wide fitness visualization of single insertions

Besides visualization of individual insertions across a genome, users can visualize multiple fitness data tracks simultaneously as well as multiple different genomes (e.g. strains) for visual inspection of the data and generating publishable figures.

3 Conclusions

MAGenTA provides complete Tn-Seq analysis from raw read processing to fitness calculations, and data visualization. MAGenTA has been used in multiple studies and enables the ability to calculate a highly quantitative measure of fitness, identify bottlenecks, perform robust statistical analyses and visually inspect the data. Moreover, we have provided example and practice files at Github as well as a detailed step-by-step manual (Supplementary Material S1) to ensure that MAGenTA is accessible to users with variable levels of bioinformatics experience.