Prediction of Oral Bioavailability in Rats: Transferring Insights from in Vitro Correlations to (Deep) Machine Learning Models Using in Silico Model Outputs and Chemical Structure Parameters.

Oral administration of drug products is a strict requirement in many medical indications. Therefore, bioavailability prediction models are of high importance for prioritization of compound candidates in the drug discovery process. However, oral exposure and bioavailability are difficult to predict, as they are the result of various highly complex factors and/or processes influenced by the physicochemical properties of a compound, such as solubility, lipophilicity, or charge state, as well as by interactions with the organism, for instance, metabolism or membrane permeation. In this study, we assess whether it is possible to predict intravenous (iv) or oral drug exposure and oral bioavailability in rats. As input parameters, we use (i) six experimentally determined in vitro and physicochemical endpoints, namely, membrane permeation, free fraction, metabolic stability, solubility, pKa value, and lipophilicity; (ii) the outputs of six in silico absorption, distribution, metabolism, and excretion models trained on the same endpoints, or (iii) the chemical structure encoded as fingerprints or simplified molecular input line entry system strings. The underlying data set for the models is an unprecedented collection of almost 1900 data points with high-quality in vivo experiments performed in rats. We find that drug exposure after iv administration can be predicted similarly well using hybrid models with in vitro- or in silico-predicted endpoints as inputs, with fold change errors (FCE) of 2.28 and 2.08, respectively. The FCEs for exposure after oral administration are higher, and here, the prediction from in vitro inputs performs significantly better in comparison to in silico-based models with FCEs of 3.49 and 2.40, respectively, most probably reflecting the higher complexity of oral bioavailability. Simplifying the prediction task to a binary alert for low oral bioavailability, based only on chemical structure, we achieve accuracy and precision close to 70%.

Refined elasticity sampling for Monte Carlo-based identification of stabilizing network patterns.

MOTIVATION: Structural kinetic modelling (SKM) is a framework to analyse whether a metabolic steady state remains stable under perturbation, without requiring detailed knowledge about individual rate equations. It provides a representation of the system's Jacobian matrix that depends solely on the network structure, steady state measurements, and the elasticities at the steady state. For a measured steady state, stability criteria can be derived by generating a large number of SKMs with randomly sampled elasticities and evaluating the resulting Jacobian matrices. The elasticity space can be analysed statistically in order to detect network positions that contribute significantly to the perturbation response. Here, we extend this approach by examining the kinetic feasibility of the elasticity combinations created during Monte Carlo sampling. RESULTS: Using a set of small example systems, we show that the majority of sampled SKMs would yield negative kinetic parameters if they were translated back into kinetic models. To overcome this problem, a simple criterion is formulated that mitigates such infeasible models. After evaluating the small example pathways, the methodology was used to study two steady states of the neuronal TCA cycle and the intrinsic mechanisms responsible for their stability or instability. The findings of the statistical elasticity analysis confirm that several elasticities are jointly coordinated to control stability and that the main source for potential instabilities are mutations in the enzyme alpha-ketoglutarate dehydrogenase.

Integration of genome-scale modeling and transcript profiling reveals metabolic pathways underlying light and temperature acclimation in Arabidopsis.

Understanding metabolic acclimation of plants to challenging environmental conditions is essential for dissecting the role of metabolic pathways in growth and survival. As stresses involve simultaneous physiological alterations across all levels of cellular organization, a comprehensive characterization of the role of metabolic pathways in acclimation necessitates integration of genome-scale models with high-throughput data. Here, we present an integrative optimization-based approach, which, by coupling a plant metabolic network model and transcriptomics data, can predict the metabolic pathways affected in a single, carefully controlled experiment. Moreover, we propose three optimization-based indices that characterize different aspects of metabolic pathway behavior in the context of the entire metabolic network. We demonstrate that the proposed approach and indices facilitate quantitative comparisons and characterization of the plant metabolic response under eight different light and/or temperature conditions. The predictions of the metabolic functions involved in metabolic acclimation of Arabidopsis thaliana to the changing conditions are in line with experimental evidence and result in a hypothesis about the role of homocysteine-to-Cys interconversion and Asn biosynthesis. The approach can also be used to reveal the role of particular metabolic pathways in other scenarios, while taking into consideration the entirety of characterized plant metabolism.

Best of both worlds: An expansion of the state of the art pKa model with data from three industrial partners.

In a unique collaboration between Simulations Plus and several industrial partners, we were able to develop a new version 11.0 of the previously published in silico pKa model, S+pKa, with considerably improved prediction accuracy. The model's training set was vastly expanded by large amounts of experimental data obtained from F. Hoffmann-La Roche AG, Genentech Inc., and the Crop Science division of Bayer AG. The previous v7.0 of S+pKa was trained on data from public sources and the Pharmaceutical division of Bayer AG. The model has shown dramatic improvements in predictive accuracy when externally validated on three new contributor compound sets. Less expected was v11.0's improvement in prediction on new compounds developed at Bayer Pharma after v7.0 was released (2013-2023), even without contributing additional data to v11.0. We illustrate chemical space coverage by chemistries encountered in the five domains, public and industrial, outline model construction, and discuss factors contributing to model's success.

A MATLAB toolbox for structural kinetic modeling.

SUMMARY: Structural kinetic modeling (SKM) enables the analysis of dynamical properties of metabolic networks solely based on topological information and experimental data. Current SKM-based experiments are hampered by the time-intensive process of assigning model parameters and choosing appropriate sampling intervals for Monte-Carlo experiments. We introduce a toolbox for the automatic and efficient construction and evaluation of structural kinetic models (SK models). Quantitative and qualitative analyses of network stability properties are performed in an automated manner. We illustrate the model building and analysis process in detailed example scripts that provide toolbox implementations of previously published literature models. AVAILABILITY: The source code is freely available for download at http://bioinformatics.uni-potsdam.de/projects/skm. CONTACT: girbig@mpimp-golm.mpg.de.

Stoichiometric capacitance reveals the theoretical capabilities of metabolic networks.

MOTIVATION: Metabolic engineering aims at modulating the capabilities of metabolic networks by changing the activity of biochemical reactions. The existing constraint-based approaches for metabolic engineering have proven useful, but are limited only to reactions catalogued in various pathway databases. RESULTS: We consider the alternative of designing synthetic strategies which can be used not only to characterize the maximum theoretically possible product yield but also to engineer networks with optimal conversion capability by using a suitable biochemically feasible reaction called 'stoichiometric capacitance'. In addition, we provide a theoretical solution for decomposing a given stoichiometric capacitance over a set of known enzymatic reactions. We determine the stoichiometric capacitance for genome-scale metabolic networks of 10 organisms from different kingdoms of life and examine its implications for the alterations in flux variability patterns. Our empirical findings suggest that the theoretical capacity of metabolic networks comes at a cost of dramatic system's changes. CONTACT: larhlimi@mpimp-golm.mpg.de, or nikoloski@mpimp-golm.mpg.de SUPPLEMENTARY INFORMATION: Supplementary tables are available at Bioinformatics online.

On the relation between reactions and complexes of (bio)chemical reaction networks.

Robustness of biochemical systems has become one of the central questions in systems biology although it is notoriously difficult to formally capture its multifaceted nature. Maintenance of normal system function depends not only on the stoichiometry of the underlying interrelated components, but also on the multitude of kinetic parameters. Invariant flux ratios, obtained within flux coupling analysis, as well as invariant complex ratios, derived within chemical reaction network theory, can characterize robust properties of a system at steady state. However, the existing formalisms for the description of these invariants do not provide full characterization as they either only focus on the flux-centric or the concentration-centric view. Here we develop a novel mathematical framework which combines both views and thereby overcomes the limitations of the classical methodologies. Our unified framework will be helpful in analyzing biologically important system properties.

The stability and robustness of metabolic states: identifying stabilizing sites in metabolic networks.

The dynamic behavior of metabolic networks is governed by numerous regulatory mechanisms, such as reversible phosphorylation, binding of allosteric effectors or temporal gene expression, by which the activity of the participating enzymes can be adjusted to the functional requirements of the cell. For most of the cellular enzymes, such regulatory mechanisms are at best qualitatively known, whereas detailed enzyme-kinetic models are lacking. To explore the possible dynamic behavior of metabolic networks in cases of lacking or incomplete enzyme-kinetic information, we present a computational approach based on structural kinetic modeling. We derive statistical measures for the relative impact of enzyme-kinetic parameters on dynamic properties (such as local stability) and apply our approach to the metabolism of human erythrocytes. Our findings show that allosteric enzyme regulation significantly enhances the stability of the network and extends its potential dynamic behavior. Moreover, our approach allows to differentiate quantitatively between metabolic states related to senescence and metabolic collapse of the human erythrocyte. We think that the proposed method represents an important intermediate step on the long way from topological network analysis to detailed kinetic modeling of complex metabolic networks.

Metabolic networks are NP-hard to reconstruct.

High-throughput data from various omics and sequencing techniques have rendered the automated metabolic network reconstruction a highly relevant problem. Our approach reflects the inherent probabilistic nature of the steps involved in metabolic network reconstruction. Here, the goal is to arrive at networks which combine probabilistic information with the possibility to obtain a small number of disconnected network constituents by reduction of a given preliminary probabilistic metabolic network. We define automated metabolic network reconstruction as an optimization problem on four-partite graph (nodes representing genes, enzymes, reactions, and metabolites) which integrates: (1) probabilistic information obtained from the existing process for metabolic reconstruction from a given genome, (2) connectedness of the raw metabolic network, and (3) clustering of components in the reconstructed metabolic network. The practical implications of our theoretical analysis refer to the quality of reconstructed metabolic networks and shed light on the problem of finding more efficient and effective methods for automated reconstruction. Our main contributions include: a completeness result for the defined problem, polynomial-time approximation algorithm, and an optimal polynomial-time algorithm for trees. Moreover, we exemplify our approach by the reconstruction of the sucrose biosynthesis pathway in Chlamydomonas reinhardtii.

A mathematical model of cocoa bean fermentation.

Cocoa bean fermentation relies on the sequential activation of several microbial populations, triggering a temporal pattern of biochemical transformations. Understanding this complex process is of tremendous importance as it is known to form the precursors of the resulting chocolate's flavour and taste. At the same time, cocoa bean fermentation is one of the least controlled processes in the food industry. Here, a quantitative model of cocoa bean fermentation is constructed based on available microbiological and biochemical knowledge. The model is formulated as a system of coupled ordinary differential equations with two distinct types of state variables: (i) metabolite concentrations of glucose, fructose, ethanol, lactic acid and acetic acid and (ii) population sizes of yeast, lactic acid bacteria and acetic acid bacteria. We demonstrate that the model can quantitatively describe existing fermentation time series and that the estimated parameters, obtained by a Bayesian framework, can be used to extract and interpret differences in environmental conditions. The proposed model is a valuable tool towards a mechanistic understanding of this complex biochemical process, and can serve as a starting point for hypothesis testing of new systemic adjustments. In addition to providing the first quantitative mathematical model of cocoa bean fermentation, the purpose of our investigation is to show how differences in estimated parameter values for two experiments allow us to deduce differences in experimental conditions.

Profiling, quantification and classification of cocoa beans based on chemometric analysis of carbohydrates using hydrophilic interaction liquid chromatography coupled to mass spectrometry.

Fifty-six cocoa bean samples from different origins and status of fermentation were analyzed by a validated hydrophilic interaction liquid chromatography-electrospray ionization-time of flight-mass spectrometry (HILIC-ESI-TOF-MS) method. The profile of the low molecular weight carbohydrate (LMWC) was analyzed by high resolution and tandem mass spectrometry, which allowed the identification of mono-, di-, tri- and tetrasaccharides, sugar alcohols and iminosugars. This study provides, for the first time in a large set of samples, a comprehensive absolute quantitative data set for the carbohydrates identified in cocoa beans (fructose, glucose, mannitol, myo-inositol, sucrose, melibiose, raffinose and stachyose). Differences in the content of carbohydrates were observed between unfermented (range of 0.9-4.9 g/g DM) and fermented (range 0.1-0.5 g/g DM) cocoa beans. The use of multivariate statistical tools allowed the identification of biomarkers suitable for cocoa bean classification according to the status of fermentation, procedure of fermentation employed and number of days of fermentation.

Degradation of cocoa proteins into oligopeptides during spontaneous fermentation of cocoa beans.

Degradation products of proteins produced during fermentation are believed to be the key precursors of a range of Maillard reactions that deliver the characteristic flavor and aroma of cocoa and chocolate. We have utilized UPLC-ESI-Q-q-TOF to identify and relatively quantify the largest collection of cocoa oligopeptides during a spontaneous fermentation time series using Ivory Coast cocoa beans. Peptides were identified, sequenced by tandem mass spectrometry and annotated based on their characteristic fragmentation pattern in the positive-ion mode. This enabled us to quantitatively trace the sequential degradation of the two main cocoa storage proteins, namely, albumin and vicilin. We observed sequential proteolytic degradation forming longer peptides in the early stages of fermentation and an increasing number of shorter peptides at the latter stages of fermentation. Protein degradation is mediated by both endo- and exopeptidases degrading at either peptide termini. In excess of 800 fermentation peptides could be unambiguously identified, providing unprecedented mechanistic details of cocoa fermentation.

Origin-based polyphenolic fingerprinting of Theobroma cacao in unfermented and fermented beans.

A comprehensive analysis of cocoa polyphenols from unfermented and fermented cocoa beans from a wide range of geographic origins was carried out to catalogue systematic differences based on their origin as well as fermentation status. This study identifies previously unknown compounds with the goal to ascertain, which of these are responsible for the largest differences between bean types. UHPLC coupled with ultra-high resolution time-of-flight mass spectrometry was employed to identify and relatively quantify various oligomeric proanthocyanidins and their glycosides amongst several other unreported compounds. A series of biomarkers allowing a clear distinction between unfermented and fermented cocoa beans and for beans of different origins were identified. The large sample set employed allowed comparison of statistically significant variations of key cocoa constituents.

Bioactivity in Rhododendron: A Systemic Analysis of Antimicrobial and Cytotoxic Activities and Their Phylogenetic and Phytochemical Origins.

The exceptional diversity of the genus Rhododendron has a strong potential for identification, characterization, and production of bioactive lead compounds for health purposes. A particularly relevant field of application is the search for new antibiotics. Here, we present a comparative analysis of nearly 90 Rhododendron species targeted toward the search for such candidate substances. Through a combination of phytochemical profiles with antimicrobial susceptibility and cytotoxicity, complemented by phylogenetic analyses, we identify seven potentially antimicrobial active but non-cytotoxic compounds in terms of mass-to-charge ratios and retention times. Exemplary bioactivity-guided fractionation for a promising Rhododendron species experimentally supports in fact one of these candidate lead compounds. By combining categorical correlation analysis with Boolean operations, we have been able to investigate the origin of bioactive effects in further detail. Intriguingly, we discovered clear indications of systems effects (synergistic interactions and functional redundancies of compounds) in the manifestation of antimicrobial activities in this plant genus.

Aseptic artificial fermentation of cocoa beans can be fashioned to replicate the peptide profile of commercial cocoa bean fermentations.

The fermentation of cocoa beans is essential for the generation of flavour precursors that are required later on to form the flavour components of chocolate. From the many different precursors that are generated, oligopeptides and free amino acids comprise a significant proportion as some of them form Maillard reaction products during the roasting process. Therefore, the diversity of peptides is an important contributing factor to the quality of a fermentation which is in turn controlled by proteolytic activity within the cocoa bean, and is driven by changes in the presence of fermentation by-products as a result of microbial activity outside the bean. Being able to control proteolytic activity within the bean using only the presence of fermentation by-products would prove a valuable tool in the study of these proteases and the processing of cocoa storage proteins. Thus, this tool would help elucidate key mechanisms that generate the components responsible for flavour. In this study, we describe an artificial fermentation system, free from microbial activity, which is able to replicate proteolytic degradation of protein as well as to generate similar peptide fragments as seen during a commercial fermentation. It was also found that acidification is a main contributor to protein degradation.

Biochemical fate of vicilin storage protein during fermentation and drying of cocoa beans.

Key cocoa-specific aroma precursors are generated during the fermentation of cocoa beans via the proteolysis of the vicilin-like globulin. Previous studies had shown that degradation of this particular 566 amino acid-long storage protein leads to three distinct subunits with different molecular masses. Although oligopeptides generated from the proteolysis of vicilin-like globulin have been studied previously, changes occurring to vicilin at different stages of fermentation have not yet been explored in detail. The aim of this study was to investigate the fate of vicilin protein from the non-fermented stage up to the dried cocoa beans. Our results showed a remarkable shift in the electrophoretic mobility of vicilin towards higher pI during the onset of fermentation. The pI-shifted subunit was found susceptible to further degradation into a lower-molecular-weight vicilin subunit. The observed pI shift correlated with, but did not depend on protein phosphorylation. Glycosylation of some but not all vicilin subunits occurred at different stages of the fermentation process. Peptides generated from vicilin throughout fermentation were analyzed by UHPLC-ESI-MS/MS revealing an initial increase and subsequent decrease in the diversity of peptides with an increasing degree of fermentation. We furthermore describe the rate of degradation of different vicilin subunits. The detected diversity and dynamics of vicilin peptides will help to define biochemical markers of distinct steps of the fermentation process.

Systematic analysis of stability patterns in plant primary metabolism.

Metabolic networks are characterized by complex interactions and regulatory mechanisms between many individual components. These interactions determine whether a steady state is stable to perturbations. Structural kinetic modeling (SKM) is a framework to analyze the stability of metabolic steady states that allows the study of the system Jacobian without requiring detailed knowledge about individual rate equations. Stability criteria can be derived by generating a large number of structural kinetic models (SK-models) with randomly sampled parameter sets and evaluating the resulting Jacobian matrices. Until now, SKM experiments applied univariate tests to detect the network components with the largest influence on stability. In this work, we present an extended SKM approach relying on supervised machine learning to detect patterns of enzyme-metabolite interactions that act together in an orchestrated manner to ensure stability. We demonstrate its application on a detailed SK-model of the Calvin-Benson cycle and connected pathways. The identified stability patterns are highly complex reflecting that changes in dynamic properties depend on concerted interactions between several network components. In total, we find more patterns that reliably ensure stability than patterns ensuring instability. This shows that the design of this system is strongly targeted towards maintaining stability. We also investigate the effect of allosteric regulators revealing that the tendency to stability is significantly increased by including experimentally determined regulatory mechanisms that have not yet been integrated into existing kinetic models.

Optimizing metabolic pathways by screening for feasible synthetic reactions.

BACKGROUND: Reconstruction of genome-scale metabolic networks has resulted in models capable of reproducing experimentally observed biomass yield/growth rates and predicting the effect of alterations in metabolism for biotechnological applications. The existing studies rely on modifying the metabolic network of an investigated organism by removing or inserting reactions taken either from evolutionary similar organisms or from databases of biochemical reactions (e.g., KEGG). A potential disadvantage of these knowledge-driven approaches is that the result is biased towards known reactions, as such approaches do not account for the possibility of including novel enzymes, together with the reactions they catalyze. RESULTS: Here, we explore the alternative of increasing biomass yield in three model organisms, namely Bacillus subtilis, Escherichia coli, and Hordeum vulgare, by applying small, chemically feasible network modifications. We use the predicted and experimentally confirmed growth rates of the wild-type networks as reference values and determine the effect of inserting mass-balanced, thermodynamically feasible reactions on predictions of growth rate by using flux balance analysis. CONCLUSIONS: While many replacements of existing reactions naturally lead to a decrease or complete loss of biomass production ability, in all three investigated organisms we find feasible modifications which facilitate a significant increase in this biological function. We focus on modifications with feasible chemical properties and a significant increase in biomass yield. The results demonstrate that small modifications are sufficient to substantially alter biomass yield in the three organisms. The method can be used to predict the effect of targeted modifications on the yield of any set of metabolites (e.g., ethanol), thus providing a computational framework for synthetic metabolic engineering.

Evolutionary significance of metabolic network properties.

Complex networks have been successfully employed to represent different levels of biological systems, ranging from gene regulation to protein-protein interactions and metabolism. Network-based research has mainly focused on identifying unifying structural properties, such as small average path length, large clustering coefficient, heavy-tail degree distribution and hierarchical organization, viewed as requirements for efficient and robust system architectures. However, for biological networks, it is unclear to what extent these properties reflect the evolutionary history of the represented systems. Here, we show that the salient structural properties of six metabolic networks from all kingdoms of life may be inherently related to the evolution and functional organization of metabolism by employing network randomization under mass balance constraints. Contrary to the results from the common Markov-chain switching algorithm, our findings suggest the evolutionary importance of the small-world hypothesis as a fundamental design principle of complex networks. The approach may help us to determine the biologically meaningful properties that result from evolutionary pressure imposed on metabolism, such as the global impact of local reaction knockouts. Moreover, the approach can be applied to test to what extent novel structural properties can be used to draw biologically meaningful hypothesis or predictions from structure alone.

Complexity of automated gene annotation.

Integration of high-throughput data with functional annotation by graph-theoretic methods has been postulated as promising way to unravel the function of unannotated genes. Here, we first review the existing graph-theoretic approaches for automated gene function annotation and classify them into two categories with respect to their relation to two instances of transductive learning on networks--with dynamic costs and with constant costs--depending on whether or not ontological relationship between functional terms is employed. The determined categories allow to characterize the computational complexity of the existing approaches and establish the relation to classical graph-theoretic problems, such as bisection and multiway cut. In addition, our results point out that the ontological form of the structured functional knowledge does not lower the complexity of the transductive learning with dynamic costs--one of the key problems in modern systems biology. The NP-hardness of automated gene annotation renders the development of heuristic or approximation algorithms a priority for additional research.