Predicting Metabolic Reaction Networks with Perturbation-Theory Machine Learning (PTML) Models

Abstract

Background: Checking the connectivity (structure) of complex Metabolic Reaction Networks (MRNs) models proposed for new microorganisms with promising properties is an important goal for chemical biology. Objective: In principle, we can perform a hand-on checking (Manual Curation). However, this is a hard task due to the high number of combinations of pairs of nodes (possible metabolic reactions). Method: In this work, we used Combinatorial, Perturbation Theory, and Machine Learning, techniques to seek a CPTML model for MRNs >40 organisms compiled by Barabasis’ group. First, we quantified the local structure of a very large set of nodes in each MRN using a new class of node index called Markov linear indices fk. Next, we calculated CPT operators for 150000 combinations of query and reference nodes of MRNs. Last, we used these CPT operators as inputs of different ML algorithms. Results: The CPTML linear model obtained using LDA algorithm is able to discriminate nodes (metabolites) with correct assignation of reactions from not correct nodes with values of accuracy, specificity, and sensitivity in the range of 85-100% in both training and external validation data series. Conclusion: Meanwhile, PTML models based on Bayesian network, J48-Decision Tree and Random Forest algorithms were identified as the three best non-linear models with accuracy greater than 97.5%. The present work opens a door to the study of MRNs of multiple organisms using PTML models.

Computational Drug Repurposing for Antituberculosis Therapy: Discovery of Multi-Strain Inhibitors

“…This means that the LQI descriptor for a molecule will have the same value regardless of the experimental condition cj used to assess the anti-TB activity of that molecule. To solve this inconvenience, we applied the adaptation of the Box–Jenkins approach, which is a distinctive characteristic of all the PTML models [ 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 55 ]: …”

“…To solve the aforementioned limitations, several researchers have emphasized the use of interpretable in silico models focused on a combination of perturbation theory concepts and machine learning techniques (PTML) [ 15 , 16 , 17 ], which can integrate different sources of chemical and biological data, enabling the simultaneous prediction of multiple biological endpoints against many targets of varying degrees of complexity. Seminal works on PTML models have found successful applications in diverse research areas such as infectious diseases [ 18 , 19 ], oncology [ 20 , 21 ], neuroscience [ 22 , 23 , 24 , 25 ], proteomics [ 26 ], metabolomics [ 27 ], nanotechnology [ 28 , 29 , 30 , 31 ], toxicology [ 32 ], and immunology and immunotoxicity [ 33 , 34 ].…”

Machine Learning Study of Metabolic NetworksvsChEMBL Data of Antibacterial Compounds

“…Novel applications of MN reconstructions of human pathogens have recently been described. These studies have focused on elucidating resistance metabolic dependencies and identifying potential drug targets and antibiotics. − The influence of the changes in MNs on the capacity of various microorganisms to survive has been demonstrated by Barabási′s team and other authors. , …”

Towards rational nanomaterial design by predicting drug–nanoparticle system interaction vs. bacterial metabolic networks

“…28 In addition, Duardo-Sanchez et al, Riera-Fernández et al, and others reported IFPTML models of MN but not included AD or NP components. [29][30][31][32] Consequently, IFPTML has been used before to solve parts of the present problem. However, there are no reports of IFPTML models including the three components AD, NP, and MN of this problem at the same time.…”