Applications of AI to problems that matter

Learn concerning the data science of AI applications to chemistry through my lay description of a recent paper

Discovering recent antibiotics to combat drug-resistant bacteria is a serious urge as bacteria change into immune to existing ones, nevertheless it is a particularly difficult and expensive endeavour. Any recent strategy to give you recent antibiotics faster and more efficiently is thus *very* welcome.

A recent study published in Nature Chemical Biology demonstrated the ability of machine learning (ML) in accelerating antibiotic discovery. The researchers used advanced algorithms to screen 1000’s of molecules and identified a promising compound called abaucin that specifically targets a pathogen called Acinetobacter baumannii which is nowadays immune to a lot of antibiotics normally employed in hospitals to treat infections—what’s called a “multidrug-resistant pathogen”. The brand new breakthrough in ML applied to antibiotic (search and) research highlights the potential of artificial intelligence in revolutionizing the sector, promising a way forward for faster, more confident, and inexpensive antibiotic development.

Conventional screening approaches for locating recent antibiotics have been limited of their effectiveness against A. baumannii because of its multidrug resistance. As I’ve touched upon a couple of times, finding recent drugs is a pressing matter:

Within the last couple of years, ML techniques began to offer novel and more efficient ways to explore chemical space via using message-passing networks, transformers, and diffusion models. Naturally, these recent approaches can increase the probabilities of finding potent antibacterial molecules. Within the study I present here, just published in Nature Chemical Biology, the researchers screened around 7,500 molecules to discover people who inhibited the expansion of A. baumannii in laboratory tests after which built a predictive ML model with which they got here up with the brand new prospective antibiotic, abaucin:

Abaucin not only demonstrates promising characteristics as a possible antibiotic, nevertheless it also exhibits selective activity against A. baumannii, making it a narrow-spectrum antibiotic with minimal effects on other bacterial species. This specificity is crucial for minimizing disruption to the body’s natural microbiota (bacteria that normally continue to exist our skin, guts, etc. and are essential to our well-being), which plays an important role in human health. Much more promising regarding its actual use as antibiotic, the study reports that abaucin effectively controls A. baumannii infections in a mouse wound model, indicating its therapeutic potential.

On the core of the work, a message-passing neural network was trained using a dataset of molecules capable (or not) of inhibiting the expansion of A. baumannii. This dataset was itself measured and reported in the identical work. Subsequently, the trained model was used to make predictions on the Drug Repurposing Hub, a comprehensive, annotated resource of FDA-approved compounds, whose purpose is to permit studies whereby already approved drugs are repurposed for brand new treatments. Here the main focus was on identifying structurally recent molecules with activity against A. baumannii. This process led to the invention of abaucin as explained above.

The ML model utilized a graph representation of the molecules and iteratively exchanged details about chemical environments around atoms through message-passing steps. The learned features were combined with fixed molecular features computed using RDKit. The datasets used for training and prediction consisted of molecules screened for growth inhibition against A. baumannii, right throughout the same work.

Key within the model is the way it converts the graph representing the structure of every molecule right into a continuous vector representation by iteratively exchanging local chemical information between adjoining atoms and bonds in a series of message-passing steps. By accumulating the vector representations of assorted local chemical regions, the model obtained a comprehensive representation of the whole compound. To enhance the learned features, fixed molecular features were computed using RDKit, one of the crucial vital libraries on the market for cheminformatics. The ultimate vector, incorporating each learned and computed features, was then used as input for a feed-forward neural network trained to predict the antibacterial properties of the molecule, as a classifier.

The dataset used for training consisted in the outcomes from a screen of seven,684 small molecules, evaluating their impact on the expansion of A. baumannii. The screening experiments resulted in 480 molecules classified as ‘energetic’ and seven,204 molecules classified as ‘inactive.’ This dataset was utilized to coach the above-described network as a binary classifier that predicted the activity of structurally recent molecules. Moreover, the Drug Repurposing Hub, containing 6,680 molecules, was employed for further predictions using an ensemble of ten classifiers.

To encode the information, the authors used SMILES strings, that are textual representations of chemical structures, after which tools from the RDKit library to interpret these SMILES strings and derive the relevant atoms and bonds. This encoding allowed the neural network to process and learn from the molecular structures effectively.

The training process involved training the message-passing neural network model on the expansion inhibition dataset using an ensemble of ten models. The hyperparameters employed included the variety of message-passing steps (3), neural network hidden size (300), variety of feed-forward layers (2), and dropout probability (0). The model’s performance was evaluated using tenfold cross-validation, a method where the dataset is split into ten subsets, and the model is trained and tested using different combos of those subsets. The chemical relationship between molecules within the training and prediction datasets was measured using Tanimoto similarity, a rating typically used to measure the similarity of two molecules -a whole topic in itself:

This study underscores the worth of ML in relevant, modern scientific research, here specifically regarding biology and antibiotic discovery that are tightly related to clinical applications. By leveraging its ability to rapidly analyze vast chemical datasets, ML enables researchers to discover molecules with targeted antibacterial properties more efficiently. This approach not only accelerates the drug discovery process but additionally increases the likelihood of finding compounds effective against highly resistant bacteria like A. baumannii.

The success of machine learning on this study opens up exciting possibilities for the long run of antibiotic research. With the continued development of advanced algorithms and computational models, scientists can optimize the means of identifying structurally diverse and functionally unique antibacterial leads. By harnessing the ability of artificial intelligence, we could also be one step closer to overcoming the worldwide challenge of antibiotic resistance.

The article:

RDKit, a library for cheminformatics widely utilized in this type of research where software must parse and manipulate molecules:

The drug repurposing hub, an open resource essential for research projects geared toward finding recent uses for existing, already-approved molecules:

A number of other cool applications of AI to chemistry and neighboring fields of science: