Link to the paper: https://pubmed.ncbi.nlm.nih.gov/35999570/
The study used deep learning to analyze metagenomics data from the International Space Station (ISS) and identified antimicrobial resistance (AMR) genes associated with cultivable strains and environmental samples.
The results revealed AMR dominance in the last flight for Kalamiella piersonii and identified antibiotic resistance genes in Enterobacter bugandensis and Bacillus cereus, which were experimentally validated.
Contributions of the paper
The paper utilized deep learning to analyze metagenomics data from the International Space Station (ISS) and identified antimicrobial resistance (AMR) genes associated with cultivable strains and environmental samples.
The study expanded the understanding of the ISS environmental microbiomes and their pathogenic potential in humans by uncovering concealed AMR determinants in metagenomics datasets.
The research provided insights into the dominance of AMR in Kalamiella piersonii and identified antibiotic resistance genes in Enterobacter bugandensis and Bacillus cereus, which were experimentally validated.
The computational predictions and validation analyses demonstrated the advantages of machine learning in characterizing AMR genes and extending the catalog of AMR genes beyond traditional cut-offs based on high DNA sequence similarity.
Practical implications of the paper
The study’s findings have practical implications for understanding and monitoring antimicrobial resistance (AMR) in space habitats, such as the International Space Station (ISS), which is crucial for long-distance space travel.
The use of deep learning algorithms to analyze metagenomics data from the ISS provides a more comprehensive understanding of the AMR genes present in cultivable strains and environmental samples. This can aid in the development of targeted strategies to mitigate AMR and ensure the safety of astronauts during space missions.
The identification of specific AMR genes in bacteria like Kalamiella piersonii, Enterobacter bugandensis, and Bacillus cereus, along with experimental validation, highlights the potential pathogenicity and resistance profiles of these microorganisms. This knowledge can inform the selection of appropriate antibiotics and treatment protocols for infections that may arise in space environments.
The study demonstrates the advantages of machine learning in uncovering concealed AMR determinants in metagenomics datasets, expanding the catalog of AMR genes, and improving our understanding of the ISS environmental microbiomes. This can contribute to the development of effective surveillance and management strategies for AMR in space habitats and beyond.
Methods used in this paper
Shotgun metagenomics data from the International Space Station (ISS) generated during the Microbial Tracking-1 (MT-1) project was analyzed using a deep learning model to identify antimicrobial resistance (AMR) genes associated with cultivable strains and environmental samples.
The paired-end 100-bp metagenomic reads were processed with Trimmomatic to trim adapter sequences and low-quality ends, and contigs were assembled using metaSPAdes. Metabat2 was used for contig binning, and CheckM was used to evaluate the quality of recovered genomes.
The genomes were annotated using Rapid Annotations using Subsystems Technology (RAST), and near identifications were predicted. Computational methods like Roary and Glimmer were used for gene comparisons and identification.
Phenotypic antibiotic resistance testing was performed using disc assays, and the resistance results were compared with interpretive charts provided by the manufacturer.
Whole-genome sequences were created using shotgun libraries and sequenced using Illumina Nextera Flex protocol. Quality control, adapter trimming, and assembly were performed using FastQC, fastp, and SPAdes, respectively. ANI calculations were done using OrthoANIu.
Data used in the paper
Shotgun metagenomics data of the International Space Station (ISS) generated during the Microbial Tracking-1 (MT-1) project, including samples collected from ISS locations and ground samples from the Crew resupply vehicle.
Whole genomes of 226 cultivable strains isolated from the MT-1 project.
21 shotgun metagenome sequences and 24 metagenome-assembled genomes (MAGs) retrieved from the ISS environmental samples.
Propidium monoazide (PMA) treated samples of viable microbes.
Antibiotic resistance profiles of two top-ranking species, Enterobacter bugandensis and Bacillus cereus, which were experimentally validated.
Note: The paper also mentions the use of computational methods and deep learning models, but the specific data used for training these models is not mentioned.
Results of the paper
The study used a deep learning model to analyze the data and identify antimicrobial resistance (AMR) genes associated with the International Space Station (ISS) surface microbiome. The analysis revealed AMR dominance in the last flight for Kalamiella piersonii, a bacteria related to urinary tract infection in humans.
Analysis of 226 pure strains isolated from the MT-1 project revealed hundreds of antibiotic resistance genes from many isolates, including two top-ranking species, Enterobacter bugandensis and Bacillus cereus. Computational predictions were experimentally validated, confirming the high resistance of these two pathogens to various beta-lactam antibiotics.
The re-analysis of short sequences and metagenome-assembled genomes (MAGs) from the ISS revealed dominance of K. piersonii antibiotic resistance in different locations of Flight 3. Phenotypic antibiotic resistance testing data obtained from traditional tests showed excellent agreement with computational predictions for the antibiotics tested.
Antibacterial Susceptibility Tests (AST) were performed for E. bugandensis and B. cereus, and the prediction patterns closely matched the AST results, demonstrating the applicability and high accuracy of computational prediction of AMR for microbiome data obtained in space.
Note: The paper also mentions providing new insights into previously unobserved antibiotic resistance classes and the limitations of the deep learning model in predicting specific compounds of antibiotics. However, these details are not directly related to the question and are not included in the answer.
We value your ideas and feel free to comment below.
Martian To Be Follow us 
