首页|期刊导航|Magnetic Medicine|A machine learning pipeline for predicting biogenic magnetic nanoparticle morphology from genoproteomic features

A machine learning pipeline for predicting biogenic magnetic nanoparticle morphology from genoproteomic featuresOA

中文摘要

Magnetotactic bacteria(MTB)biomineralize intracellular,membrane-bound magnetic nanoparticles known as magnetosomes crystals of magnetite(Fe_(3)O_(4))or greigite(Fe_(3)S_(4))that exhibit exceptional uniformity,bio-compatibility,and magnetic precision.The morphology of these nanocrystals(cuboctahedral,prismatic,or bullet-shaped)critically determines their magnetic anisotropy,coercivity,and collective behavior,influencing performance in biomedical applications such as magnetic resonance imaging(MRI),magnetic hyperthermia,and targeted drug delivery.However,the genetic and proteomic determinants governing magnetosome shape remain poorly understood.In this study,we present a machine learning pipeline that predicts magnetosome morphology directly from bacterial genoproteomic features.Using curated datasets of magnetotactic bacterial genomes and proteomes,the random forest model achieved a mean accuracy of 98%in cross-validation.Feature-importance analysis revealed a hierarchical biomineralization model:broad morphological classes are defined by gene cluster presence(e.g.,mad),while subtle physicochemical variations in key proteins-especially MamL and MamO fine-tune final crystal geometry.Comparative analyses identified statistically significant differences in aromaticity(p=0.0065)and isoelectric point(p=0.0195)of MamO between prismatic and cuboctahedral morphotypes,implicating aromatic residues and electrostatic effects in crystal elongation.The pipeline is open,modular,and designed for accessibility,enabling automated in silico screening of MTB strains for desired na-noparticle morphologies.This approach bridges computational biology,microbial genomics,and magnetic na-nomaterials science,offering a scalable framework for rational design of biogenic magnetic nanoparticles tai-lored for magnetogenetics,hyperthermia therapy,and advanced diagnostic applications.

Oksana Gorobets;Kyryl Khakhno;Svitlana Gorobets

National Technical University of Ukraine"Kyiv Polytechnic Institute",Faculty of Physics and Mathematics,Kyiv 03056,UkraineNational Technical University of Ukraine"Kyiv Polytechnic Institute",Faculty of Biotechnology and Biotechnics,Kyiv 03056,UkraineNational Technical University of Ukraine"Kyiv Polytechnic Institute",Faculty of Biotechnology and Biotechnics,Kyiv 03056,Ukraine

生物科学

Magnetotactic bacteriaMagnetosomeBiogenic magnetic nanoparticlesMachine learningGenoproteomicsBiomineralizationMorphology prediction,synthetic biology

《Magnetic Medicine》 2026 (1)

P.22-33,12

10.1016/j.magmed.2026.100063

评论