A master’s thesis from UNED has allowed the development of a artificial intelligence system based on neural networks identifies the Achilles heel of bacteria, that is, the genes responsible for their resistance to antibiotics.
Everything indicates that in the coming years the number of people in space will multiply. Therefore, in addition to having technological means to safely remove them from our atmosphere, to various destinations such as the Moon, the International Space Station or, in the somewhat more distant future, Mars, it is also necessary to guarantee health in an environment as hostile as space.
It is currently known that prolonged stays on the International Space Station have consequences on the immune systemwhich makes astronauts especially vulnerable, especially if one takes into account that once in space, the chances of returning to Earth at the right time to treat a disease are very slim or practically nil.
For this reason, space agencies are investing a lot of research time in reducing the chances of being affected by a disease while away from Earth. One of your priorities is reduce bacterial resistancethat is, the ability of bacteria to defend themselves against antibiotics.
What has UNED contributed to research on bacterial resistance in space?
In this area, the final master’s project carried out at the UNED by Pedro Madrigal, who is currently a researcher at the European Bioinformatics Institute (EMBL-EBI), has meant a very important leap in quality, because in it A formula based on artificial intelligence and deep learning is described, which improves the identification of genes from the microbiome of the International Space Station.he.
“The International Space Station has a distribution of microorganisms, or microbiome, on its surfaces that is characteristic of this habitat, but this is not surprising.. For example, using metagenomics to obtain genome sequences of different microorganisms in metro stations around the world, differences of the same magnitude have been discovered”, explains Pedro Madrigal. But you have to know it and study it well.
The objective of this research gestated in the Master in Research in Artificial Intelligence from UNED has been understand how microbial diversity and its damage affect spaceflightas well as helping researchers and astronauts address the risk of infection and antibiotic resistance.
And for this they have used machine learning algorithms, which is the branch of artificial intelligence dedicated to automating part of the scientific methodin which computers observe data to build a model based on them and at the same time use this model as a hypothesis to be validated, which provides solutions to complex problems.
How the genetic resistance of bacteria is deciphered
Current sequencing techniques allow direct access and profiling of the full set of metagenomic DNAwhere the genes responsible for antibiotic resistance (known as ARGs) are usually identified or predicted based on the “best hits” of sequence searches in existing databases.
Unfortunately, this approach produces a high false negative rate (genes identified as non-resistant that are). To address the limitations of this orientation, a supervised deep learning approach (Deep learning), specifically DeepARG developed by other researchers in 2018.
Specifically, they were used DeepARG-SS and DeepARG-LS deep learning models, for short read sequences and full-length sequences of the gene, respectively, and that have shown high sensitivity for the detection of antibiotic resistance genes. It is a deep neural network (deep learning) that considers a dissimilarity matrix between the different categories of ARG genes. This dissimilarity matrix represents the difference between two determined ARGs. The output layer of the neural network in DeepARG is composed of 30 units that correspond to the different categories of antibiotic resistance.
How the experiments were done
The results of the last flight revealed the domain antibiotic resistant genes of Kalamiella piersonii, a bacterium associated with urinary tract infection in humans. In the analysis we used 226 inbred strains isolated from the MT-1 project hundreds of antibiotic resistance genes were detectedincluding two high-ranking species that corresponded to strains of Enterobacter bugandensis and Bacillus cereus.
The computational predictions were experimentally validated by antibiotic resistance profiles in these two species, and a high degree of concordance between both techniques was demonstrated.. Specifically, the disk assay data confirmed the high resistance of these two pathogens to several beta-lactam antibiotics, which are the most common group of antibiotics.
The professors of the Department of Artificial Intelligence at UNED, Elena Gaudioso and Félix Hernández del Olmo, together with Afshin Beheshti (researcher at NASA Ames Research Center), have been in charge of directing this work, supervising the formulation of the Deep algorithms learning (a particular type of Machine Learning) that have been used, its application to the data provided by the institutions that are part of the project and the validation of the results obtained.
The next step, explains Pedro Madrigal, does not necessarily involve designing new drugs, although it is true that there are companies trying to facilitate this through microgravity. “The next important issues for space agencies regarding antimicrobial resistance are, on the one hand, achieve a system that allows studying and monitoring the evolution of antimicrobial resistance during long-duration manned flights, if possible in real time”. On the other hand, the “optimal design of an astropharmacy for these missions”where antibiotics on board must cover a wide range of potential threats to the crew.
It has been proven that radiation from space increases mutations in bacteria, which end up generating resistance to antibiotics.
And if that is added to the fact that the immune system of astronauts is altered during space flight, you have “the perfect storm.” Hence the importance of this study and others in the same line to get to know the bacteria much better.
