Computational protocol: Alternative Evolutionary Paths to Bacterial Antibiotic Resistance Cause Distinct Collateral Effects

Similar protocols

Protocol publication

[…] We employed an established pipeline encoded in serial bash and Perl scripts used previously for the genomic analysis of P. aeruginosa PA14 (). Briefly, reads with unreliable quality were removed using Skewer (). Samples were then mapped to the published P. aeruginosa_UCBPP_PA14_NC008463 reference genome available at (http://pseudomonas.com/strain/download; last accessed May 19, 2017). Mapping was performed using bwa and samtools (; ) and then visually inspected for low-quality areas using IGV (Integrated genome viewer, Broad Institute; www.broadinstitute.org/software/igv/; last accessed May 19, 2017).Duplicated regions were removed for single nucleotide polymorphisms and structural variant calling (SNPs and SV) using MarkDuplicates in Picardtools (https://braod.institute.github.io/picard/; last accessed May 19, 2017). To call SNPs and small indels above a threshold frequency of 0.1 and base quality above 20, we employed both frequentist and heuristic methods using respectively SNVer and VarScan (; ). To identify larger indels and other SV, we used Pindel and CNVnator (; ). The resulting output files were filtered for duplicates, ancestral variants, and variants found in the evolved controls. We used a combination of sources to identify and annotate the variants using snpEff (), DAVID, the Pseudomonas database (available online at: http://pseudomonas.com; last accessed May 19, 2017) and information from published work. Further count statistics, analysis and visualizations were done in the R platform (R Core Team).Mutational diversity was calculated as in (). Briefly, we calculated the entropy H= −∑[pj( log⁡2pj + (1−pj log⁡2(1−pj)], where pj is the probability that a given locus j is mutated in a random population. H then measures the diversity of mutated loci in the populations adapted to a given drug. Standard error was obtained from jackknife resampling in the R platform.In order to link the observed collateral effects to the underlying genetic changes we performed a hierarchical clustering analysis. For this, we focused on four treatments, which repeatedly produced contrasting patterns of collateral effects. These included populations adapted to either GEN or STR (the two aminoglycosides), which produced variation in their collateral profiles towards PIT and CAR. We also considered the reverse two cases, for which replicated populations that had adapted to PIT and CAR showed contrasting patterns of collateral effects towards GEN and STR. For these four cases, we first obtained the Euclidean similarity of the sensitivities of evolved populations against the considered drugs. Then we used hierarchical clustering based on Ward’s minimum variance method, including the Ward’s criterion, which aims at finding compact, spherical clusters, and combined it with bootstrapping to asses cluster stability (). The same process was then used to infer clusters based on the genomic profiles of the same populations, including only genes with mutations within their coding regions. For each antibiotic, we then built dendograms for the clustering results and assessed to what extent given genomic clusters coincided with clusters having collateral resistance or sensitivity phenotypes.The obtained genome sequences are available from NCBI SRA database under the BioProject number: PRJNA355367. […]

Pipeline specifications

Software tools skewer, BWA, SAMtools, IGV, Picard, SNVer, VarScan, Pindel, CNVnator, SnpEff
Application Genome data visualization
Organisms Bacteria, Homo sapiens, Pseudomonas aeruginosa
Diseases Pseudomonas Infections
Chemicals Aminoglycosides, Gentamicins, Penicillins