Machine learning models reveal new drug design opportunities

Machine Learning


This article has been reviewed in accordance with Science X's editorial processes and policies. The editors have highlighted the following attributes while ensuring the authenticity of the content:

fact confirmed

Peer-reviewed publications

trusted sources

proofread


Basic relationships between descriptors. Each data point represents a molecule and is projected into a two-dimensional space of two descriptors as shown in the figure. Some of the most common properties in descriptors are: be linear correlation, b nonlinear correlation, and c Uncorrelated. The numbers in each panel are calculated by standard correlation (Pearson coefficient). CIjrank correlation, RIj. As shown, the rank correlation better captures the nonlinear relationship shown in the middle panel. credit: communication chemistry (2024). DOI: 10.1038/s42004-024-01161-y

× close


Basic relationships between descriptors. Each data point represents a molecule and is projected into a two-dimensional space of two descriptors as shown in the figure. Some of the most common properties in descriptors are: be linear correlation, b nonlinear correlation, and c Uncorrelated. The numbers in each panel are calculated by standard correlation (Pearson coefficient). CIjrank correlation, RIj. As shown, the rank correlation better captures the nonlinear relationship shown in the middle panel. credit: communication chemistry (2024). DOI: 10.1038/s42004-024-01161-y

Pathogens are nothing without their ability to adapt, and their ability to protect themselves from antibiotics is of increasing public health concern. A research team led by Los Alamos National Laboratory is using machine learning, an application of artificial intelligence, to discover new types of antibiotics, especially among pathogens deemed important by the World Health Organization because they are highly bacterial. identified molecular characteristics that may lead to resistance.

Research results will be published in a magazine communication chemistry.

“Some pathogens have properties that make them very effective at resisting antibiotics,” says Los Alamos scientist Gnana Gnanakaran. “The discovery of specific compounds that can penetrate and inhibit some pathogens is a needle in a haystack because of the vast heterogeneity and depth of chemical space and the complexity of molecular interactions across bacterial membranes.” It's a challenging challenge. The approach we take allows us to explore the bacterial-specific molecular profiles needed for successful drug development.”

Bacterial defense against antibiotics

Gram-negative bacteria have an outer membrane that makes it difficult for the compounds that make up antibiotics to penetrate, and the bacteria may inhibit the effectiveness of antibiotics by expelling compounds that accidentally get inside.

Data-driven models have the potential to identify molecular properties that can overcome these bacterial defenses, but the precise calculations to make these decisions are difficult and require extensive computing resources. Masu. Chemically diverse compounds can contain many related properties. Machine learning-driven research has reduced the relevant spectrum of these properties and established rules of thumb to predict a compound's ability to penetrate the outer membrane of bacteria.

Machine learning model identifies pathogen-fighting properties

The research team focused specifically on the Gram-negative bacterium Pseudomonas aeruginosa to identify relevant descriptors associated with compounds and predict whether those compounds can penetrate the outer membrane of the bacterium and avoid excretion. developed a machine learning model. Using Los Alamos' high-performance computing capabilities, the team extracted the molecular characteristics of permeation from simulations of his 1,260 chemically diverse compounds passing through bacterial membranes.

Their analysis sheds new light on the critical properties required for drug candidates to effectively penetrate P. aeruginosa and opens the door to similar data-based studies in other Gram-negative pathogens.

“The machine learning techniques we employed in this analysis represent a promising approach for similar data-driven studies in other biological membranes, including the blood-brain barrier,” Gnanakaran said.

For more information:
Pedro D. Manrique et al. Predicting the permeation of compounds across the outer membrane of Pseudomonas aeruginosa using molecular descriptors, communication chemistry (2024). DOI: 10.1038/s42004-024-01161-y

Magazine information:
communication chemistry



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *