Machine learning and structural analysis of Mycobacterium tuberculosis pan-genome identifies genetic signatures of antibiotic resistance

Kavvas, Erol; Catoiu, Edward; Mih, Nathan; Yurkovich, James T.; Seif, Yara; Dillon, Nicholas; Heckmann, David; Anand, Amitesh; Yang, Laurence T.; Nizet, Victor; Monk, Jonathan M.

doi:10.1038/s41467-018-06634-y

Cited by 132 publications

(125 citation statements)

References 62 publications

(50 reference statements)

Supporting

Mentioning

121

Contrasting

Order By: Relevance

“…These mutations occur in non-AMR genes in order to accommodate the potentially reduced fitness cost of maintaining the primary AMR conferring genes or SNPs [67][68][69][70] . The results of this study, and previous AI studies focusing on AMR 25,27 , suggest that these changes could be widespread throughout the genome. Indeed, although each model had a few genes with very high total feature importance values, most genes were contributing k-mers to the decision trees for each model.…”

Section: Discussionsupporting

confidence: 68%

“…The average very major error rate (VME), which is defined as resistant genomes that are erroneously predicted to be susceptible, and the average major error rate (ME), which is defined as susceptible genomes that are erroneously predicted to be resistant, tend to go down as gene set size increases. Although the core gene set models described in Figure 1 have lower F1 scores and higher error rates than full-genome models that have been published previously [21][22][23][24]27,29 , their accuracies are striking given the small sizes of the input data sets and the removal of well-annotated AMR genes.…”

Section: Amr Models Based On Core Genes Have Predictive Powermentioning

confidence: 84%

“…These predictions are typically made using either rules-based or machine learning models [16][17][18][19][20] . Several studies have also built machine learning models for predicting AMR phenotypes by using assembled genomes or pan genomes as training sets [21][22][23][24][25][26][27] . In these cases, the machine learning algorithm detects the most discriminating features (typically short nucleotide k-mers) from a training set with laboratory-derived AMR phenotypes.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Predicting Antimicrobial Resistance Using Conserved Genes

Nguyen

Olson

Shukla

et al. 2020

Preprint

View full text Add to dashboard Cite

Abbreviations AIartificial intelligence CI confidence interval ME major error, susceptible genomes predicted to be resistant MIC minimum inhibitory concentration PATRIC Pathosystems resource integration center PLF PATRIC local protein family RAST Rapid annotation using subsystem technology RF Random Forest SR susceptible and resistant VME very major error, resistant genomes predicted to be susceptible XGB XGBoost Abstract A growing number of studies have shown that machine learning algorithms can be used to accurately predict antimicrobial resistance (AMR) phenotypes from bacterial sequence data. In these studies, models are typically trained using input features derived from comprehensive sets of known AMR genes or whole genome sequences. However, it can be difficult to determine whether genomes and their corresponding sets of AMR genes are complete when sequencing contaminated or metagenomic samples. In this study, we explore the possibility of using incomplete genome sequence data to predict AMR phenotypes. Machine learning models were built from randomly-selected sets of core genes that are held in common among the members of a species, and the AMR-conferring genes were removed based on their protein annotations. For Klebsiella pneumoniae, Mycobacterium tuberculosis, Salmonella enterica, and Staphylococcus aureus, we report that it is possible to classify susceptible and resistant phenotypes with average F1 scores ranging from 0.80-0.89 with as few as 100 conserved non-AMR genes, with very major error rates ranging from 0.11-0.23 and major error rates ranging from 0.10-0.20. Models built from core genes have predictive power in the cases where the primary AMR mechanism results from SNPs or horizontal gene transfer. By randomly sampling non-overlapping sets of core genes for use in these models, we show that F1 scores and error rates are stable and have little variance between replicates. Potential biases from strainspecific SNPs, phylogenetic sampling, and imbalances in the phylogenetic distribution of susceptible and resistant strains do not appear to have an impact on this result. Although these small core gene models have lower accuracies and higher error rates than models built from the corresponding assembled genomes, the results suggest that sufficient variation exists in the core non-AMR genes of a species for predicting AMR phenotypes. Overall this study suggests that building models from conserved genes may be a potentially useful strategy for predicting AMR phenotypes when genomes are incomplete.

show abstract

Section: Discussionsupporting

confidence: 68%

Section: Amr Models Based On Core Genes Have Predictive Powermentioning

confidence: 84%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Predicting Antimicrobial Resistance Using Conserved Genes

Nguyen

Olson

Shukla

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Databases such as the Pathosystems Resource Integration Center (PATRIC) (Wattam et al ; Davis et al ; Wattam et al ) and the Antibiotic Resistance Genes Database (ARDB) (Liu and Pop ) currently provide genomes with AR metadata. In some recent studies (Her and Wu ; Kavvas et al ; Moradigaravand et al ), machine learning techniques were applied to find resistance to varieties of antibiotics from whole genome sequences using databases of known AR genes. These studies first constructed a pan‐genome with identifying core and accessory gene clusters, then computed features such as presence–absence patterns and the population structures of gene clusters, and finally applied a machine learning algorithm to identity putative AR genes using the extracted features.…”

Section: Introductionmentioning

confidence: 99%

Capreomycin resistance prediction in two species of Mycobacterium using a stacked ensemble method

2019

View full text Add to dashboard Cite

Aims Predicting bacterial resistance provides valuable information that can assist in clinical decisions. With recent advances in whole genome sequencing technology, the detection of antibiotic resistance (AR) proteins directly from genomic data is becoming feasible. AR genes/proteins can be identified using best‐hit methods that work by comparing candidate sequences with known AR genes in public databases. However, these approaches may fail to detect resistance genes with sequences that differ significantly from known sequences. Our goal is to develop a machine learning technique to accurately predict capreomycin resistance in Mycobacteria with low false discovery rates. Methods and Results We present a stacked ensemble learning model as an alternative to traditional DNA sequence alignment‐based methods using optimal features generated from the physicochemical, evolutionary and secondary structure properties of protein sequences. We train logistic regression, C5.0 and support vector machine (SVM) algorithms as our base classifiers, and our stacked ensemble predictors combine the results from the base classifiers to achieve higher accuracy. Compared with our most accurate base classifier (SVM), our most accurate stacked ensemble predictor increases training accuracy by 2·43%. Our stacked ensemble predictors achieve test accuracy up to 81·25%. Conclusions We developed a stacked ensemble model to predict capreomycin resistance for Mycobacteria with an accuracy >80% using protein sequences with sequence similarity ranging between 10% and 70%. This performance cannot be achieved with best‐hit methods due to differences in sequence similarity. Significance and Impact of the Study Today an estimated one‐half million cases of multidrug‐resistant (MDR) and extensively drug‐resistant (XDR) tuberculosis (TB) occur annually worldwide at a great cost. Because capreomycin is a second‐line drug used to treat drug‐resistant TB, the ability to use a machine learning approach to classify capreomycin‐resistant TB in a timely manner is crucial for the successful treatment of MDR or XDR TB.

show abstract

“…(12) Compounding the general problem of antibiotic resistance are challenges in developing and approving new antibiotics. (13) This has spurred research into understanding resistance mechanisms, (14,15) host defences, (16)(17)(18) diagnostics (19) and antibiotic generation. (20,21) Ultimately, without a rapid and perhaps general method to develop new targeted antibiotics, therapies might relapse towards those of the pre-antibiotic era.…”

Section: Introductionmentioning

confidence: 99%

Antibody-Recruiting Protein-Catalyzed Capture Agents to Combat Antibiotic-Resistant Bacteria

Idso

Akhade

Arrieta-Ortiz

et al. 2019

Preprint

View full text Add to dashboard Cite

Antibiotic resistant infections are projected to cause over 10 million deaths by 2050, yet the development of new antibiotics has slowed. This points to an urgent need for methodologies for the rapid development of antibiotics against emerging drug resistant pathogens. We report on a generalizable combined computational and synthetic approach, called antibody-recruiting proteincatalyzed capture agents (AR-PCCs), to address this challenge. We applied the combinatorial PCC technology to identify macrocyclic peptide ligands against highly conserved surface protein epitopes of carbapenem-resistant Klebsiella pneumoniae, an opportunistic gram-negative pathogen with drug resistant strains. Multi-omic data combined with bioinformatic analyses identified epitopes of the highly expressed MrkA surface protein of K. pneumoniae for targeting in PCC screens. The top-performing ligand exhibited high-affinity (EC50~50 nM) to full-length MrkA, and selectively bound to MrkAexpressing K. pneumoniae, but not to other pathogenic bacterial species. AR-PCCs conjugated with immunogens promoted antibody recruitment to K. pneumoniae, leading to phagocytosis and phagocytic killing by macrophages. The rapid development of this highly targeted antibiotic implies that the integrated computational and synthetic toolkit described here can be used for the accelerated production of antibiotics against drug resistant bacteria. resistant K. pneumoniae, and that the basic technology might provide a route towards drugging "undruggable" pathogenic bacteria. Results and DiscussionMulti-omic analyses to select target a protein on K. pneumoniae Here we describe the algorithm used to identify protein targets, and epitopes on those targets, for drugging K. pneumoniae using AR-PCCs. Traditional drugging strategies tend to rely on disrupting the function of, for example, an enzyme by competing for occupancy within a strategic hydrophobic binding pocket. Our requirements are very different. Instead, favorable aspects of target proteins are high expression levels on only the pathogen of interest, plus localization of that protein to the outer membrane or extracellular space of the pathogen. Further, once such a target protein is identified, there are additional considerations regarding which epitopes of that protein present the greatest opportunities for exploiting AR-PCCs. The flow diagram in Figure 2A delineates the strategy for identifying MrkA as an ideal target protein. Protein expression levels can vary across environments and growth phases, so we analyzed the transcriptional data reported by Guilhen et al(40) to identify proteins

show abstract

Machine learning and structural analysis of Mycobacterium tuberculosis pan-genome identifies genetic signatures of antibiotic resistance

Cited by 132 publications

References 62 publications

Predicting Antimicrobial Resistance Using Conserved Genes

Predicting Antimicrobial Resistance Using Conserved Genes

Capreomycin resistance prediction in two species of Mycobacterium using a stacked ensemble method

Antibody-Recruiting Protein-Catalyzed Capture Agents to Combat Antibiotic-Resistant Bacteria

Contact Info

Product

Resources

About