Causal Artificial Intelligence Models of Food Quality Data.

Research background: The aim of this study is to emphasize the importance of artificial intelligence (AI) and causality modelling of food quality and analysis with 'big data'. AI with structural causal modelling (SCM), based on Bayesian networks and deep learning, enables the integration of theoretical field knowledge in food technology with process production, physicochemical analytics and consumer organoleptic assessments. Food products have complex nature and data are highly dimensional, with intricate interrelations (correlations) that are difficult to relate to consumer sensory perception of food quality. Standard regression modelling techniques such as multiple ordinary least squares (OLS) and partial least squares (PLS) are effectively applied for the prediction by linear interpolations of observed data under cross-sectional stationary conditions. Upgrading linear regression models by machine learning (ML) accounts for nonlinear relations and reveals functional patterns, but is prone to confounding and failed predictions under unobserved nonstationary conditions. Confounding of data variables is the main obstacle to applications of the regression models in food innovations under previously untrained conditions. Hence, this manuscript focuses on applying causal graphical models with Bayesian networks to infer causal relationships and intervention effects between process variables and consumer sensory assessment of food quality. Experimental approach: This study is based on the data available in the literature on the process of wheat bread baking quality, consumer sensory quality assessments of fermented milk products, and professional wine tasting data. The data for wheat baking quality were regularized by the least absolute shrinkage and selection operator (LASSO elastic net). Bayesian statistics was applied for the evaluation of the model joint probability function for inferring the network structure and parameters. The obtained SCMs are presented as directed acyclic graphs (DAG). D-separation criteria were applied to block confounding effects in estimating direct and total causal effects of process variables and consumer perception on food quality. Probability distributions of causal effects of the intervention of individual process variables on quality are presented as partial dependency plots determined by Bayesian neural networks. In the case of wine quality causality, the total causal effects determined by SCMs are positively validated by the double machine learning (DML) algorithm. Results and conclusions: The data set of 45 continuous variables corresponding to different chemical, physical and biochemical variables of wheat properties from seven Croatian cultivars during two years of controlled cultivation were analysed. LASSO regularization of the data set yielded the ten key predictors, accounting for 98 % variance of the baking quality data. Based on the key variables, the quality predictive random forest model with 75 % cross-validation accuracy was derived. Causal analysis between the quality and key predictors was based on the Bayesian model shown as a DAG graph. Protein content shows the most important direct causal effect with the corresponding path coefficient of 0.71, and THMM (total high-molecular-mass glutenin subunits) content was an indirect cause with a path coefficient of 0.42, and protein total average causal effect (ACE) was 0.65. The large data set of the quality of fermented milk products included binary consumer sensory data (taste, odour, turbidity), continuous physical variables (temperature, fat, pH, colour) and three grade classes of products by consumer quality assessment. A random forest model was derived for the prediction of the quality classification with an out-of-bag (OOB) error of 0.28 %. The Bayesian network model predicts that the direct causes of the taste classification are temperature, colour and fat content, while the direct causes of the quality classification are temperature, turbidity, odour and fat content. The key quality grade ACE of temperature -0.04 grade/°C and 0.3 quality grade/fat content were estimated. The temperature ACE dependency shows a nonlinear type as negative saturation with the 'breaking' point at 60 °C, while for fat ACE had a positive linear trend. Causal quality analysis of red and white wine was based on the large data set of eleven continuous variables of physical and chemical properties and quality assessments classified in ten classes, from 1 to 10. Each classification was obtained in triplicate by a panel of professional wine tasters. A non-structural double machine learning (DML) algorithm was applied for total ACE quality assessment. The alcohol content of red and white wine had the key positive ACE relative factor of 0.35 quality/alcohol, while volatile acidity had the key negative ACE of -0.2 quality/acidity. The obtained ACE predictions by the unstructured DML algorithm are in close agreement with the ACE obtained by the structural SCM. Novelty and scientific contribution: Novel methodologies and results for the application of causal artificial intelligence models in the analysis of consumer assessment of the quality of food products are presented. The application of Bayesian network structural causal models (SCM) enables the d-separation of pronounced effects of confounding between parameters in noncausal regression models. Based on the SCM, inference of ACE provides substantiated and validated research hypotheses for new products and support for decisions of potential interventions for improvement in product design, new process introduction, process control, management and marketing.

Application of Causality Modelling for Prediction of Molecular Properties for Textile Dyes Degradation by LPMO.

The textile industry is one of the largest water-polluting industries in the world. Due to an increased application of chromophores and a more frequent presence in wastewaters, the need for an ecologically favorable dye degradation process emerged. To predict the decolorization rate of textile dyes with Lytic polysaccharide monooxygenase (LPMO), we developed, validated, and utilized the molecular descriptor structural causality model (SCM) based on the decision tree algorithm (DTM). Combining mathematical models and theories with decolorization experiments, we have elucidated the most important molecular properties of the dyes and confirm the accuracy of SCM model results. Besides the potential utilization of the developed model in the treatment of textile dye-containing wastewater, the model is a good base for the prediction of the molecular properties of the molecule. This is important for selecting chromophores as the reagents in determining LPMO activities. Dyes with azo- or triarylmethane groups are good candidates for colorimetric LPMO assays and the determination of LPMO activity. An adequate methodology for the LPMO activity determination is an important step in the characterization of LPMO properties. Therefore, the SCM/DTM model validated with the 59 dyes molecules is a powerful tool in the selection of adequate chromophores as reagents in the LPMO activity determination and it could reduce experimentation in the screening experiments.

Current state and prospects of biotechnology in Central and Eastern European countries. Part II: new and preaccession EU countries(CRO, RO, B&H, SRB).

Innovation holds the potential for economic prosperity. Biotechnology (BT) has proved to be a viable vehicle for the development and utilization of technologies, which has brought not only advances to society, but also career opportunities to nation-states that have enabling conditions. In this review, we assess the current state of BT-related activities within selected new and preaccession EU countries (NPA) of CEE region namely Croatia, Romania, Bosnia and Herzegovina and Serbia, examining educational programs, research activity, enterprises, and the financing systems. The field of BT covers a broad area of activities, including medical, food and agriculture, aquaculture or marine, environmental, biofuels, bioinformatics, and many others. Under the European Commission (EC), member-states are to set their Research and Innovation Strategies for Smart Specialization (RIS3), to identify priorities or strengths in order to develop knowledge intensive economies. As the four countries highlighted in this review are in the early stages of implementing RIS3 or have not yet fully formulated, it presents an opportunity to learn from the successes and failures of those that have already received major structural funds from the EC. A critical point will be the ability of the public and private sectors' actors to align, in the implementation of RIS3 as new investment instruments emerge, and to concentrate efforts on a few select target goals, rather than distribute funding widely without respect to a long-term vision.

Reliability of a decision-tree model in predicting occupational lead poisoning in a group of highly exposed workers.

OBJECTIVE: This study aimed to provide the toxicological profile of some lead-exposed workers and obtain a predictive model for lead poisoning. METHODS: Data regarding external and absorbed exposure were collected from 585 subjects employed in ten metallurgical production departments. Airborne lead concentration, blood lead level (BLL), cumulative blood lead index (CBLI), urine delta-aminolevulinic acid (DALA), age, workplace/section, exposure period, and whether reported lead poisoning as occupational disease were examined using ANOVA, and, post-ANOVA, Pearson correlation matrix, PCA (principal component analysis), decision-tree modeling, and logistic modeling. RESULTS: BLL was less sensitive than CBLI in predicting poisoning. Decision-tree modeling highlighted the importance of CBLI ≥1,041 µg.years/dl and air lead concentration ≥0.3 mg/m(3) in the occurrence of occupational poisoning. Age ≥48 years and DALA ≥19.3 mg/L were also factors. CONCLUSIONS: Workers were at risk of poisoning as a result of their long term unacceptable exposure. Decision-tree modeling is potentially useful for risk management. Am. J. Ind. Med. 59:575-582, 2016. © 2016 Wiley Periodicals, Inc.

Near-Infrared Spectroscopic Analysis of Total Phenolic Content and Antioxidant Activity of Berry Fruits.

This study evaluates the feasibility of using near-infrared (NIR) spectroscopy as a rapid and environmentally friendly technique for validation and prediction of the total phenolic content (TPC) and antioxidant activity (AOA) indices (as 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging, inhibition time (IT) of the Briggs-Rauscher oscillating reaction, and relative antioxidant capacity (RAC)) of berry fruit extracts. The analysed berry samples originated from Croatia (blackberries, wild blueberries, raspberries, red currants and strawberries) and Bulgaria (wild blueberries, raspberries and strawberries). Principal component analysis and partial least squares (PLS) regression were used from the set of chemometric tools in distinguishing and validating the measured berry fruit extract. ANOVA and PCA showed no significant impact of the origin and freshness of the samples. PLS models were developed to validate the relationship of NIR spectra with TPC and AOA of berry fruits. Representativeness of the models was expressed with the R2 and the ratio of performance to deviation. Calculated R2 values were above 0.84 and the ratio of performance to deviation was between 1.8 and 3.1, indicating adequacy of the PLS models.

Catabolism of terbuthylazine by mixed bacterial culture originating from s-triazine-contaminated soil.

The s-triazine herbicide terbuthylazine (TERB) has been used as the main substitute of atrazine in many EU countries for more than 10 years. However, the ecological consequences of this substitution are still not fully understood. Since the fate of triazine herbicides is primarily dependent on microbial degradation, in this paper, we investigated the ability of a mixed bacterial culture, M3-T, originating from s-triazine-contaminated soil, to degrade TERB in liquid culture and soil microcosms. The M3-T culture grown in mineral medium with TERB as the N source and citrate as the C source degraded 50 mg L(-1) of TERB within 3 days of incubation. The culture was capable of degrading TERB as the sole C and N source, though at slower degradation kinetics. A thorough LC-MS analysis of the biodegradation media showed the formation of hydroxyterbuthylazine (TERB-OH) and N-t-butylammelide (TBA) as major metabolites, and desethylterbuthylazine (DET), hydroxydesethylterbuthylazine (DET-OH) and cyanuric acid (CA) as minor metabolites in the TERB degradation pathway. TBA was identified as a bottleneck in the catabolic pathway leading to its transient accumulation in culture media. The supplementation of glucose as the exogenous C source had no effect on TBA degradation, whereas citrate inhibited its disappearance. The addition of M3-T to sterile soil artificially contaminated with TERB at 3 mg kg(-1) of soil resulted in an accelerated TERB degradation with t 1/2 value being about 40 times shorter than that achieved by the native microbial community. Catabolic versatility of M3-T culture makes it a promising seed culture for accelerating biotransformation processes in s-triazine-contaminated environment.

New approach to the evaluation of lignocellulose derived by-products impact on lytic-polysaccharide monooxygenase activity by using molecular descriptor structural causality model.

Lytic-polysaccharide monooxygenase (LPMO) is one of the most important enzyme involved in biocatalytic lignocellulose degradation, and therefore inhibition of LPMO has significant effects on all related processes. Structural causality model (SCM) were established to evaluate impact of phenolic by-products in lignocellulose hydrolysates on LPMO activity. The molecular descriptors GATS4c, ATS2m, BIC3 and VR2_Dzs were found to be significant in describing inhibition. The causalities of the molecular descriptors and LPMO activity are determined by evaluating the directed acyclic graph (DAG) and the d-separation algorithm. The maximum causality for LPMO activation is ß = 0.79 by BIC3 and the maximum causality of inhibition is ß = -0.56 for the GATS4c descriptor. The model has the potential to predict the inhibition of LPMO and its application could be useful in selecting an appropriate lignocellulose pretreatment method to minimise the production of a potent inhibitor. This will subsequently lead to more efficient lignocellulose degradation process.

Lin-log Model of E. coli Central Metabolism.

Mathematical models of dynamics of metabolic pathways are used for analysis of complex regulations of biochemical reactions as an intrinsic property of a metabolism. The models are derived under assumptions of kinetic rate functions and usually result in simplification in view of the model theoretical scope and/or its practical application. The main obstacle in kinetic modeling is the dimensionality of the parametric space, its nonlinearity and ill-conditioned relations for kinetic parameter estimation. In this work these problems are effectively resolved by use of an approximate linear-logarithmic (Lin-log) applied in analysis of regulation of Escherichia coli central metabolism. Complex multiplicative Michaelis-Menten kinetic rate expressions are transformed into simple in parameter linear functions and non-linear logarithmic dependencies on concentrations of substrates, and cofactors. The Lin-log kinetic rates enable direct estimation of rate elasticities which are the key parameters in metabolic control analysis (MCA). Due to in the parameter linearity, the estimation problem is solved in a non-iterative least square algorithm. Applied is singular value decomposition (SVD) algorithm for system matrix pseudoinversion with the eigenvalue cut-off threshold at 0.01. The results are presented as parameters of enzyme activities and reaction elasticities. Evaluated activities and elasticities provided insight into the fluxes regulation. Comparison of the simulation results by Lin-log and Michaelis-Menten model reveals that errors are of the same order of magnitude.

Effect of different variables on the efficiency of the Baker's yeast cell disruption process to obtain alcohol dehydrogenase activity.

Cell disruption process of dry baker's yeast was studied in this work to obtain maximum activity of alcohol dehydrogenase (ADH). Disruption by ultrasonication, glass beads, and combination of these two methods was compared. A 1.8-fold increase of ADH activity can be achieved by combining glass beads with ultrasonication in comparison to ultrasonication. To achieve maximum volume activity of ADH, the effect of different variables on the cell disruption process was investigated (time, glass bead diameter, mass of glass beads, and ultrasound amplitude). Using the Design-Expert© software, 24 factorial experimental design was performed. Two ultrasound probes were tested: MS 73 and KE 76. Optimal conditions (process variables) for cell disruption process were obtained. Optimal ADH activities after cell disruption with MS 73 and KE 76 probes were 1,890.9 and 1,531.7 U cm⁻³, respectively. Necessary ultrasonication time and ultrasound amplitude should be at the maximum values in the investigated variable range (30 min and 62 %). Bead size should be at maximum (4 mm) when using MS 73 probe and at minimum (0.3 mm) when using KE 76 probe. Partial purification of the enzyme was carried out and it was kinetically characterized using several oxidation and reduction systems.