In a recent review published in Science, researchers discussed the role of artificial intelligence (AI) in preventing outbreaks of infectious diseases and future pandemics.
Study: Leveraging artificial intelligence in the fight against infectious diseases. Image Credit: SomYuZu/Shutterstock.com
Despite developments in molecular genetics, computers, and pharmaceutical chemistry, infectious diseases remain a serious global health problem.
Multidisciplinary cooperation will be required to address the difficulties posed by disease outbreaks, pandemics, and antibiotic resistance.
In combination with synthetic and systems biology, AI is accelerating progress, increasing anti-infective medication discovery, improving our comprehension of infection biology, and expediting diagnostic research.
About the review
In the present review, researchers presented the challenges in preventing infectious illnesses and the contribution of AI to disease prevention.
Global Challenges in preventing infectious diseases
Challenges in understanding the biological mechanisms underlying disease and developing infection prevention measures are essential in managing outbreaks and new pathogens like severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), monkeypox, Ebola, H5N1 influenza, Marburg virus, Zika, measles, MERS, and Escherichia coli.
Problematic pathogenic organisms include methicillin-resistant Staphylococcus aureus (MRSA), carbapenem-resistant Enterobacteriaceae (CRE), vancomycin-resistant Enterococcus (VRE), multidrug-resistant tuberculosis (MDR-TB), extended-spectrum β-lactamase (ESBL)-secreting bacterial organisms, and persistent pathogens such as Neisseria gonorrhoeae, Candida auris, T. gondii, and P. falciparum.
Antimicrobial stewardship, developing novel anti-infective medications, and understanding their mechanisms of action are crucial to combat these challenges.
Additionally, developing low-cost and field-deployable diagnostics, improving test accuracy, detecting antimicrobial resistance, and making effective disease treatments available are essential for addressing persistent and neglected infections [such as Lyme disease, chronic hepatitis B virus (HBV) and hepatitis C virus (HCV) infections, chronic mycotic infections, human immunodeficiency virus (HIV)-caused acquired immunodeficiency syndrome (AIDS), and those among individuals with poor access to health resources].
Artificial intelligence and machine learning for preventing infections
AI-based approaches have the potential to integrate large amounts of quantitative and omics data, making them particularly adept at dealing with biological complexity.
Machine learning (ML), a subcategory of artificial intelligence, uses data to train machines to make predictions and has helped facilitate searches of small-molecule databases.
ML approaches include supervised graph neural networks and unsupervised generative models. Supervised machine learning algorithms examine structured and unstructured glycan, protein, nucleic acid, and cell phenotypic information to uncover important characteristics and molecular structures that regulate interactions between hosts, pathogens, and immune system responses.
Inverse vaccinology, which predicts antigens based on immunological and genetic data, has been aided by supervised ML techniques like Vaxign-ML.
De novo chemical arrangements and peptide chains are proposed using generative ML models, which may be produced and assessed. Drug development can also be aided by generative systems such as GPT-4 and NVIDIA's BioNeMo, which integrate different scientific data streams to improve our comprehension of the fundamental biological and chemical dynamics.
AI can predict anti-infective medication activity, drug-target interactions, and therapeutic design. ML approaches to anti-infective drug discovery have centered on training models to identify new drugs or uses of currently used drugs. A key benefit of ML approaches is that they can virtually screen compound libraries at a scale (>109 compounds) that would be impossible to screen empirically.
AI approaches relevant to anti-infective drug discovery include inputs such as phenotypic screens, target-specific screens, and anti-infective susceptibilities; models including graph neural networks, random forest classifiers, and explainable models; and outputs such as growth inhibition, antimicrobial activity, and target-binding activity.
Recent advancements in merging artificial intelligence with artificial biology, genetic expression analysis, imaging, and mass spectrometry have significantly increased our capacity to identify infections and predict medication resistance.
AI models are used for gene expression-, mass spectrometry-, and imaging-based diagnostics, and AST remains important for informing the use of anti-infective drugs. AI can elucidate infection biology, facilitate vaccine design, and inform anti-infective treatment strategies.
Infection biology inputs include macromolecular sequences, protein structures, microscopy, and morphology; models include network modeling, interaction modeling, and language modeling; and outputs include immunogenicity, inter-protein interactions, and pathogen killing and escape.
For vaccine design, AI inputs include nucleic acid or protein sequences, protein structures, and antigen-binding information; models include sequence-to-function-type models, classifier ensembles, and neural networks; and outputs include antigen presentation, vaccine efficacy, and translational efficacy.
Due to the strong programmability of biological components, the regular synthesis of big or sequence-based information sets, and the capacity of ML to retrieve pertinent data from biological and molecular systems in disease biological sciences, AI can aid synthetic biology studies and the development of diagnostics.
Based on the review findings, ML and AI have revolutionized infectious disease research by analyzing large datasets and providing valuable insights. However, challenges in diagnosis include low data quality, limited generalizability, and high diagnostic predictions.
Experiments involving large datasets and comprehensive benchmarking datasets are required to improve ML models.
Multi-dimensional drug interaction prediction can enhance treatment options, forecast side effects, and boost the success rates of novel medications in clinical research.