Limits...
ARTIST: high-resolution genome-wide assessment of fitness using transposon-insertion sequencing.

Pritchard JR, Chao MC, Abel S, Davis BM, Baranowski C, Zhang YJ, Rubin EJ, Waldor MK - PLoS Genet. (2014)

Bottom Line: ARTIST uses simulation-based normalization to model and compensate for experimental noise, and thereby enhances the statistical power in conditional TIS analyses.ARTIST also employs a novel adaptation of the hidden Markov model to generate statistically robust, high-resolution, annotation-independent maps of fitness-linked loci across the entire genome.Using ARTIST, we sensitively and comprehensively define Mycobacterium tuberculosis and Vibrio cholerae loci required for host infection while limiting inclusion of false positive loci.

View Article: PubMed Central - PubMed

Affiliation: Department of Microbiology, Harvard Medical School, Boston, Massachusetts, United States of America; Department of Immunology and Infectious Disease, Harvard School of Public Health, Boston, Massachusetts, United States of America.

ABSTRACT
Transposon-insertion sequencing (TIS) is a powerful approach for deciphering genetic requirements for bacterial growth in different conditions, as it enables simultaneous genome-wide analysis of the fitness of thousands of mutants. However, current methods for comparative analysis of TIS data do not adjust for stochastic experimental variation between datasets and are limited to interrogation of annotated genomic elements. Here, we present ARTIST, an accessible TIS analysis pipeline for identifying essential regions that are required for growth under optimal conditions as well as conditionally essential loci that participate in survival only under specific conditions. ARTIST uses simulation-based normalization to model and compensate for experimental noise, and thereby enhances the statistical power in conditional TIS analyses. ARTIST also employs a novel adaptation of the hidden Markov model to generate statistically robust, high-resolution, annotation-independent maps of fitness-linked loci across the entire genome. Using ARTIST, we sensitively and comprehensively define Mycobacterium tuberculosis and Vibrio cholerae loci required for host infection while limiting inclusion of false positive loci. ARTIST is applicable to a broad range of organisms and will facilitate TIS-based dissection of pathways required for microbial growth and survival under a multitude of conditions.

Show MeSH

Related in: MedlinePlus

Con-ARTIST improves detection of conditionally essential genes compared to previous studies.(A) Overlap of the conditionally essential genes in M. tuberculosis required for mouse infection as determined by Con-ARTIST, Sassetti et al. [15] and Zhang et al. [7]. The overlap between Con-ARTIST and Sassetti et al. was significantly better than the overlap between Zhang et al. and Sassetti et al. (p-val<0.05, one-tailed Fisher's exact test). (B) The standard deviations of p-values across 100 simulation-based MWU tests were plotted for either Con-ARTIST conditionally essential genes in M. tuberculosis that overlap with Zhang et al. (75 genes), or for genes that were significant only in the Zhang et al. dataset (267 genes). In all three mice, genes that overlap between Con-ARTIST and Zhang et al. had significantly (*, p-value<0.0005) narrower ranges of p-value standard deviations across MWU tests than genes categorized as conditionally essential only by Zhang et al. (C) Overlap of conditionally essential genes required for V. cholerae rabbit infection as defined by Con-ARTIST, Kamp et al. [16] or Fu et al. [11]. Genes that were defined as defective for in vitro growth by Kamp et al. were filtered from both the Con-ARTIST and Fu et al. results. The overlap between Con-ARTIST and Kamp et al. was significantly higher than the overlap between Kamp et al. and Fu et al. (p-value<0.0001 by one-tailed Fisher's exact test). (D) In-frame deletions were constructed in several V. cholerae genes that were either identified as required for rabbit infection in previous studies (but not by Con-ARTIST) or unique predicted to be conditionally essential (CE) in this study. WT and mutant cells were first barcoded and then pooled to infect rabbits. The ratios of each mutant barcode compared to WT sequences were compared before and after infection to generate a competitive index. The competitive indexes for each mutant are coordinately colored according to the individual animal from which they were derived. One outlier measurement was identified (in Δvc0432) using the Grubbs test and removed. *, significantly underrepresented, p-value≤0.01 by Kruskal-Wallis test, with Dunn's test for multiple comparisons.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4222735&req=5

pgen-1004782-g004: Con-ARTIST improves detection of conditionally essential genes compared to previous studies.(A) Overlap of the conditionally essential genes in M. tuberculosis required for mouse infection as determined by Con-ARTIST, Sassetti et al. [15] and Zhang et al. [7]. The overlap between Con-ARTIST and Sassetti et al. was significantly better than the overlap between Zhang et al. and Sassetti et al. (p-val<0.05, one-tailed Fisher's exact test). (B) The standard deviations of p-values across 100 simulation-based MWU tests were plotted for either Con-ARTIST conditionally essential genes in M. tuberculosis that overlap with Zhang et al. (75 genes), or for genes that were significant only in the Zhang et al. dataset (267 genes). In all three mice, genes that overlap between Con-ARTIST and Zhang et al. had significantly (*, p-value<0.0005) narrower ranges of p-value standard deviations across MWU tests than genes categorized as conditionally essential only by Zhang et al. (C) Overlap of conditionally essential genes required for V. cholerae rabbit infection as defined by Con-ARTIST, Kamp et al. [16] or Fu et al. [11]. Genes that were defined as defective for in vitro growth by Kamp et al. were filtered from both the Con-ARTIST and Fu et al. results. The overlap between Con-ARTIST and Kamp et al. was significantly higher than the overlap between Kamp et al. and Fu et al. (p-value<0.0001 by one-tailed Fisher's exact test). (D) In-frame deletions were constructed in several V. cholerae genes that were either identified as required for rabbit infection in previous studies (but not by Con-ARTIST) or unique predicted to be conditionally essential (CE) in this study. WT and mutant cells were first barcoded and then pooled to infect rabbits. The ratios of each mutant barcode compared to WT sequences were compared before and after infection to generate a competitive index. The competitive indexes for each mutant are coordinately colored according to the individual animal from which they were derived. One outlier measurement was identified (in Δvc0432) using the Grubbs test and removed. *, significantly underrepresented, p-value≤0.01 by Kruskal-Wallis test, with Dunn's test for multiple comparisons.

Mentions: To further assess Con-ARTIST's utility, we compared M. tuberculosis and V. cholerae genes classified by Con-ARTIST as required for optimal in vivo growth to those identified in previously published TIS or microarray-based studies [7], [11], [15], [16]. For M. tuberculosis, the Con-ARTIST pipeline classified 118 genes (Table S1) as conditionally essential for mouse infection with high likelihood (all insertion sites within these genes have>90% probability of being conditionally essential. Most of these genes (84 or 71%) overlap with those identified in previous studies by Zhang et al. [7] or Sassetti et al. [15] (Figure 4A). This overlap is significantly greater (p-value<0.05 by one-sided Fisher's exact test) than the overlaps for the Sassetti et al. and Zhang et al. datasets, which were 36% and 28%, respectively. Additionally, in all three mice, the p-values for Con-ARTIST's conditionally essential genes that overlap with those of Zhang et al. had significantly lower standard deviations across the simulation-based MWU tests than did the p-values for genes that were identified only in Zhang et al. (Figure 4B), which includes a much larger set of conditionally essential genes (371), the majority of which were not found either by our analysis or the Sassetti et al. microarray study.


ARTIST: high-resolution genome-wide assessment of fitness using transposon-insertion sequencing.

Pritchard JR, Chao MC, Abel S, Davis BM, Baranowski C, Zhang YJ, Rubin EJ, Waldor MK - PLoS Genet. (2014)

Con-ARTIST improves detection of conditionally essential genes compared to previous studies.(A) Overlap of the conditionally essential genes in M. tuberculosis required for mouse infection as determined by Con-ARTIST, Sassetti et al. [15] and Zhang et al. [7]. The overlap between Con-ARTIST and Sassetti et al. was significantly better than the overlap between Zhang et al. and Sassetti et al. (p-val<0.05, one-tailed Fisher's exact test). (B) The standard deviations of p-values across 100 simulation-based MWU tests were plotted for either Con-ARTIST conditionally essential genes in M. tuberculosis that overlap with Zhang et al. (75 genes), or for genes that were significant only in the Zhang et al. dataset (267 genes). In all three mice, genes that overlap between Con-ARTIST and Zhang et al. had significantly (*, p-value<0.0005) narrower ranges of p-value standard deviations across MWU tests than genes categorized as conditionally essential only by Zhang et al. (C) Overlap of conditionally essential genes required for V. cholerae rabbit infection as defined by Con-ARTIST, Kamp et al. [16] or Fu et al. [11]. Genes that were defined as defective for in vitro growth by Kamp et al. were filtered from both the Con-ARTIST and Fu et al. results. The overlap between Con-ARTIST and Kamp et al. was significantly higher than the overlap between Kamp et al. and Fu et al. (p-value<0.0001 by one-tailed Fisher's exact test). (D) In-frame deletions were constructed in several V. cholerae genes that were either identified as required for rabbit infection in previous studies (but not by Con-ARTIST) or unique predicted to be conditionally essential (CE) in this study. WT and mutant cells were first barcoded and then pooled to infect rabbits. The ratios of each mutant barcode compared to WT sequences were compared before and after infection to generate a competitive index. The competitive indexes for each mutant are coordinately colored according to the individual animal from which they were derived. One outlier measurement was identified (in Δvc0432) using the Grubbs test and removed. *, significantly underrepresented, p-value≤0.01 by Kruskal-Wallis test, with Dunn's test for multiple comparisons.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4222735&req=5

pgen-1004782-g004: Con-ARTIST improves detection of conditionally essential genes compared to previous studies.(A) Overlap of the conditionally essential genes in M. tuberculosis required for mouse infection as determined by Con-ARTIST, Sassetti et al. [15] and Zhang et al. [7]. The overlap between Con-ARTIST and Sassetti et al. was significantly better than the overlap between Zhang et al. and Sassetti et al. (p-val<0.05, one-tailed Fisher's exact test). (B) The standard deviations of p-values across 100 simulation-based MWU tests were plotted for either Con-ARTIST conditionally essential genes in M. tuberculosis that overlap with Zhang et al. (75 genes), or for genes that were significant only in the Zhang et al. dataset (267 genes). In all three mice, genes that overlap between Con-ARTIST and Zhang et al. had significantly (*, p-value<0.0005) narrower ranges of p-value standard deviations across MWU tests than genes categorized as conditionally essential only by Zhang et al. (C) Overlap of conditionally essential genes required for V. cholerae rabbit infection as defined by Con-ARTIST, Kamp et al. [16] or Fu et al. [11]. Genes that were defined as defective for in vitro growth by Kamp et al. were filtered from both the Con-ARTIST and Fu et al. results. The overlap between Con-ARTIST and Kamp et al. was significantly higher than the overlap between Kamp et al. and Fu et al. (p-value<0.0001 by one-tailed Fisher's exact test). (D) In-frame deletions were constructed in several V. cholerae genes that were either identified as required for rabbit infection in previous studies (but not by Con-ARTIST) or unique predicted to be conditionally essential (CE) in this study. WT and mutant cells were first barcoded and then pooled to infect rabbits. The ratios of each mutant barcode compared to WT sequences were compared before and after infection to generate a competitive index. The competitive indexes for each mutant are coordinately colored according to the individual animal from which they were derived. One outlier measurement was identified (in Δvc0432) using the Grubbs test and removed. *, significantly underrepresented, p-value≤0.01 by Kruskal-Wallis test, with Dunn's test for multiple comparisons.
Mentions: To further assess Con-ARTIST's utility, we compared M. tuberculosis and V. cholerae genes classified by Con-ARTIST as required for optimal in vivo growth to those identified in previously published TIS or microarray-based studies [7], [11], [15], [16]. For M. tuberculosis, the Con-ARTIST pipeline classified 118 genes (Table S1) as conditionally essential for mouse infection with high likelihood (all insertion sites within these genes have>90% probability of being conditionally essential. Most of these genes (84 or 71%) overlap with those identified in previous studies by Zhang et al. [7] or Sassetti et al. [15] (Figure 4A). This overlap is significantly greater (p-value<0.05 by one-sided Fisher's exact test) than the overlaps for the Sassetti et al. and Zhang et al. datasets, which were 36% and 28%, respectively. Additionally, in all three mice, the p-values for Con-ARTIST's conditionally essential genes that overlap with those of Zhang et al. had significantly lower standard deviations across the simulation-based MWU tests than did the p-values for genes that were identified only in Zhang et al. (Figure 4B), which includes a much larger set of conditionally essential genes (371), the majority of which were not found either by our analysis or the Sassetti et al. microarray study.

Bottom Line: ARTIST uses simulation-based normalization to model and compensate for experimental noise, and thereby enhances the statistical power in conditional TIS analyses.ARTIST also employs a novel adaptation of the hidden Markov model to generate statistically robust, high-resolution, annotation-independent maps of fitness-linked loci across the entire genome.Using ARTIST, we sensitively and comprehensively define Mycobacterium tuberculosis and Vibrio cholerae loci required for host infection while limiting inclusion of false positive loci.

View Article: PubMed Central - PubMed

Affiliation: Department of Microbiology, Harvard Medical School, Boston, Massachusetts, United States of America; Department of Immunology and Infectious Disease, Harvard School of Public Health, Boston, Massachusetts, United States of America.

ABSTRACT
Transposon-insertion sequencing (TIS) is a powerful approach for deciphering genetic requirements for bacterial growth in different conditions, as it enables simultaneous genome-wide analysis of the fitness of thousands of mutants. However, current methods for comparative analysis of TIS data do not adjust for stochastic experimental variation between datasets and are limited to interrogation of annotated genomic elements. Here, we present ARTIST, an accessible TIS analysis pipeline for identifying essential regions that are required for growth under optimal conditions as well as conditionally essential loci that participate in survival only under specific conditions. ARTIST uses simulation-based normalization to model and compensate for experimental noise, and thereby enhances the statistical power in conditional TIS analyses. ARTIST also employs a novel adaptation of the hidden Markov model to generate statistically robust, high-resolution, annotation-independent maps of fitness-linked loci across the entire genome. Using ARTIST, we sensitively and comprehensively define Mycobacterium tuberculosis and Vibrio cholerae loci required for host infection while limiting inclusion of false positive loci. ARTIST is applicable to a broad range of organisms and will facilitate TIS-based dissection of pathways required for microbial growth and survival under a multitude of conditions.

Show MeSH
Related in: MedlinePlus