Causality, machine learning and human insight.

Modern instruments generate BIG DATA that require information extraction before they can be used. A hybrid modelling framework for that is presented and illustrated. Its purpose is to convert meaningless data to meaningful information and to contribute to a theoretical, practical, and democratic basis for tomorrow's handling of BIG DATA in science and technology.

Corrigendum: Hyperspectral Video Analysis by Motion and Intensity Preprocessing and Subspace Autoencoding.

[This corrects the article DOI: 10.3389/fchem.2022.818974.].

Hyperspectral Video Analysis by Motion and Intensity Preprocessing and Subspace Autoencoding.

Hyperspectral imaging has recently gained increasing attention from academic and industrial world due to its capability of providing both spatial and physico-chemical information about the investigated objects. While this analytical approach is experiencing a substantial success and diffusion in very disparate scenarios, far less exploited is the possibility of collecting sequences of hyperspectral images over time for monitoring dynamic scenes. This trend is mainly justified by the fact that these so-called hyperspectral videos usually result in BIG DATA sets, requiring TBs of computer memory to be both stored and processed. Clearly, standard chemometric techniques do need to be somehow adapted or expanded to be capable of dealing with such massive amounts of information. In addition, hyperspectral video data are often affected by many different sources of variations in sample chemistry (for example, light absorption effects) and sample physics (light scattering effects) as well as by systematic errors (associated, e.g., to fluctuations in the behaviour of the light source and/or of the camera). Therefore, identifying, disentangling and interpreting all these distinct sources of information represents undoubtedly a challenging task. In view of all these aspects, the present work describes a multivariate hybrid modelling framework for the analysis of hyperspectral videos, which involves spatial, spectral and temporal parametrisations of both known and unknown chemical and physical phenomena underlying complex real-world systems. Such a framework encompasses three different computational steps: 1) motions ongoing within the inspected scene are estimated by optical flow analysis and compensated through IDLE modelling; 2) chemical variations are quantified and separated from physical variations by means of Extended Multiplicative Signal Correction (EMSC); 3) the resulting light scattering and light absorption data are subjected to the On-The-Fly Processing and summarised spectrally, spatially and over time. The developed methodology was here tested on a near-infrared hyperspectral video of a piece of wood undergoing drying. It led to a significant reduction of the size of the original measurements recorded and, at the same time, provided valuable information about systematic variations generated by the phenomena behind the monitored process.

Genome-wide association mapping for milk fat composition and fine mapping of a QTL for de novo synthesis of milk fatty acids on bovine chromosome 13.

BACKGROUND: Bovine milk is widely regarded as a nutritious food source for humans, although the effects of individual fatty acids on human health is a subject of debate. Based on the assumption that genomic selection offers potential to improve milk fat composition, there is strong interest to understand more about the genetic factors that influence the biosynthesis of bovine milk and the molecular mechanisms that regulate milk fat synthesis and secretion. For this reason, the work reported here aimed at identifying genetic variants that affect milk fatty acid composition in Norwegian Red cattle. Milk fatty acid composition was predicted from the nation-wide recording scheme using Fourier transform infrared spectroscopy data and applied to estimate heritabilities for 36 individual and combined fatty acid traits. The recordings were used to generate daughter yield deviations that were first applied in a genome-wide association (GWAS) study with 17,343 markers to identify quantitative trait loci (QTL) affecting fatty acid composition, and next on high-density and sequence-level datasets to fine-map the most significant QTL on BTA13 (BTA for Bos taurus chromosome). RESULTS: The initial GWAS revealed 200 significant associations, with the strongest signals on BTA1, 13 and 15. The BTA13 QTL highlighted a strong functional candidate gene for de novo synthesis of short- and medium-chained saturated fatty acids; acyl-CoA synthetase short-chain family member 2. However, subsequent fine-mapping using single nucleotide polymorphisms (SNPs) from a high-density chip and variants detected by resequencing showed that the effect was more likely caused by a second nearby gene; nuclear receptor coactivator 6 (NCOA6). These findings were confirmed with results from haplotype studies. NCOA6 is a nuclear receptor that interacts with transcription factors such as PPARγ, which is a major regulator of bovine milk fat synthesis. CONCLUSIONS: An initial GWAS revealed a highly significant QTL for de novo-synthesized fatty acids on BTA13 and was followed by fine-mapping of the QTL within NCOA6. The most significant SNPs were either synonymous or situated in introns; more research is needed to uncover the underlying causal DNA variation(s).

Frequency Analysis and Feature Reduction Method for Prediction of Cerebral Palsy in Young Infants.

The aim of this paper is to achieve a model for prediction of cerebral palsy based on motion data of young infants. The prediction is formulated as a classification problem to assign each of the infants to one of the healthy or with cerebral palsy groups. Unlike formerly proposed features that are mostly defined in the time domain, this study proposes a set of features derived from frequency analysis of infants' motions. Since cerebral palsy affects the variability of the motions, and frequency analysis is an intuitive way of studying variability, suggested features are suitable and consistent with the nature of the condition. In the current application, a well-known problem, few subjects and many features, was initially encountered. In such a case, most classifiers get trapped in a suboptimal model and, consequently, fail to provide sufficient prediction accuracy. To solve this problem, a feature selection method that determines features with significant predictive ability is proposed. The feature selection method decreases the risk of false discovery and, therefore, the prediction model is more likely to be valid and generalizable for future use. A detailed study is performed on the proposed features and the feature selection method: the classification results confirm their applicability. Achieved sensitivity of 86%, specificity of 92% and accuracy of 91% are comparable with state-of-the-art clinical and expert-based methods for predicting cerebral palsy.

High-throughput biochemical fingerprinting of Saccharomyces cerevisiae by Fourier transform infrared spectroscopy.

Single-channel optical density measurements of population growth are the dominant large scale phenotyping methodology for bridging the gene-function gap in yeast. However, a substantial amount of the genetic variation induced by single allele, single gene or double gene knock-out technologies fail to manifest in detectable growth phenotypes under conditions readily testable in the laboratory. Thus, new high-throughput phenotyping technologies capable of providing information about molecular level consequences of genetic variation are sorely needed. Here we report a protocol for high-throughput Fourier transform infrared spectroscopy (FTIR) measuring biochemical fingerprints of yeast strains. It includes high-throughput cultivation for FTIR spectroscopy, FTIR measurements and spectral pre-treatment to increase measurement accuracy. We demonstrate its capacity to distinguish not only yeast genera, species and populations, but also strains that differ only by a single gene, its excellent signal-to-noise ratio and its relative robustness to measurement bias. Finally, we illustrated its applicability by determining the FTIR signatures of all viable Saccharomyces cerevisiae single gene knock-outs corresponding to lipid biosynthesis genes. Many of the examined knock-out strains showed distinct, highly reproducible FTIR phenotypes despite having no detectable growth phenotype. These phenotypes were confirmed by conventional lipid analysis and could be linked to specific changes in lipid composition. We conclude that the introduced protocol is robust to noise and bias, possible to apply on a very large scale, and capable of generating biologically meaningful biochemical fingerprints that are strain specific, even when strains lack detectable growth phenotypes. Thus, it has a substantial potential for application in the molecular functionalization of the yeast genome.

150 years of the mass action law.

This year we celebrate the 150th anniversary of the law of mass action. This law is often assumed to have been "there" forever, but it has its own history, background, and a definite starting point. The law has had an impact on chemistry, biochemistry, biomathematics, and systems biology that is difficult to overestimate. It is easily recognized that it is the direct basis for computational enzyme kinetics, ecological systems models, and models for the spread of diseases. The article reviews the explicit and implicit role of the law of mass action in systems biology and reveals how the original, more general formulation of the law emerged one hundred years later ab initio as a very general, canonical representation of biological processes.

Frequency-based features for early cerebral palsy prediction.

In this paper we aim at predicting cerebral palsy, the most serious and lifelong motor function disorder in children, at an early age by analysing infants' motion data. An essential step for doing so is to extract informative features with high class separability. We propose a set of features derived from frequency analysis of the motion data. Then, we evaluate the practicality of our features on one of the richest data sets collected to study this disease. In this data set, the motion data are extracted from both electromagnetic sensors as well as video camera. The proposed features are used for classifying both data sets. Using these features, we manage to achieve promising classification performance. Classification accuracy of 91% for the sensor data and 88% for the video-derived data show not only the advantage of employing these features for predicting cerebral palsy, but also that replacing electromagnetic sensors with a video camera is feasible.

A computational pipeline for quantification of mouse myocardial stiffness parameters.

The mouse is an important model for theoretical-experimental cardiac research, and biophysically based whole organ models of the mouse heart are now within reach. However, the passive material properties of mouse myocardium have not been much studied. We present an experimental setup and associated computational pipeline to quantify these stiffness properties. A mouse heart was excised and the left ventricle experimentally inflated from 0 to 1.44kPa in eleven steps, and the resulting deformation was estimated by echocardiography and speckle tracking. An in silico counterpart to this experiment was built using finite element methods and data on ventricular tissue microstructure from diffusion tensor MRI. This model assumed a hyperelastic, transversely isotropic material law to describe the force-deformation relationship, and was simulated for many parameter scenarios, covering the relevant range of parameter space. To identify well-fitting parameter scenarios, we compared experimental and simulated outcomes across the whole range of pressures, based partly on gross phenotypes (volume, elastic energy, and short- and long-axis diameter), and partly on node positions in the geometrical mesh. This identified a narrow region of experimentally compatible values of the material parameters. Estimation turned out to be more precise when based on changes in gross phenotypes, compared to the prevailing practice of using displacements of the material points. We conclude that the presented experimental setup and computational pipeline is a viable method that deserves wider application.

Emulating facial biomechanics using multivariate partial least squares surrogate models.

A detailed biomechanical model of the human face driven by a network of muscles is a useful tool in relating the muscle activities to facial deformations. However, lengthy computational times often hinder its applications in practical settings. The objective of this study is to replace precise but computationally demanding biomechanical model by a much faster multivariate meta-model (surrogate model), such that a significant speedup (to real-time interactive speed) can be achieved. Using a multilevel fractional factorial design, the parameter space of the biomechanical system was probed from a set of sample points chosen to satisfy maximal rank optimality and volume filling. The input-output relationship at these sampled points was then statistically emulated using linear and nonlinear, cross-validated, partial least squares regression models. It was demonstrated that these surrogate models can mimic facial biomechanics efficiently and reliably in real-time.

Estimating muscle activation patterns using a surrogate model of facial biomechanics.

Analyzing the muscle activities that drive the expressive facial gestures can be a useful tool in assessing one's emotional state of mind. Since the skin motion is much easier to measure in comparison to the actual electrical excitation signal of facial muscles, a biomechanical model of the human face driven by these muscles can be a useful tool in relating the geometric information to the muscle activity. However, long computational time often hinders its practicality. The objective of this study was to replace the precise but computationally demanding biomechanical model by a much faster multivariate meta-model (surrogate model), such that a significant speedup (real-time interactive speed) can be achieved and data from the biomechanical model can be practically exploited. Using the proposed surrogate, muscle activation patterns of six key facial expressions were estimated in the iterative fit from the structured-light scanned geometric information.

Estimation of ion competition via correlated responsivity offset in linear ion trap mass spectrometry analysis: theory and practical use in the analysis of cyanobacterial hepatotoxin microcystin-LR in extracts of food additives.

Responsivity is a conversion qualification of a measurement device given by the functional dependence between the input and output quantities. A concentration-response-dependent calibration curve represents the most simple experiment for the measurement of responsivity in mass spectrometry. The cyanobacterial hepatotoxin microcystin-LR content in complex biological matrices of food additives was chosen as a model example of a typical problem. The calibration curves for pure microcystin and its mixtures with extracts of green alga and fish meat were reconstructed from the series of measurement. A novel approach for the quantitative estimation of ion competition in ESI is proposed in this paper. We define the correlated responsivity offset in the intensity values using the approximation of minimal correlation given by the matrix to the target mass values of the analyte. The estimation of the matrix influence enables the approximation of the position of a priori unknown responsivity and was easily evaluated using a simple algorithm. The method itself is directly derived from the basic attributes of the theory of measurements. There is sufficient agreement between the theoretical and experimental values. However, some theoretical issues are discussed to avoid misinterpretations and excessive expectations.

Multi-way metamodelling facilitates insight into the complex input-output maps of nonlinear dynamic models.

BACKGROUND: Statistical approaches to describing the behaviour, including the complex relationships between input parameters and model outputs, of nonlinear dynamic models (referred to as metamodelling) are gaining more and more acceptance as a means for sensitivity analysis and to reduce computational demand. Understanding such input-output maps is necessary for efficient model construction and validation. Multi-way metamodelling provides the opportunity to retain the block-wise structure of the temporal data typically generated by dynamic models throughout the analysis. Furthermore, a cluster-based approach to regional metamodelling allows description of highly nonlinear input-output relationships, revealing additional patterns of covariation. RESULTS: By presenting the N-way Hierarchical Cluster-based Partial Least Squares Regression (N-way HC-PLSR) method, we here combine multi-way analysis with regional cluster-based metamodelling, together making a powerful methodology for extensive exploration of the input-output maps of complex dynamic models. We illustrate the potential of the N-way HC-PLSR by applying it both to predict model outputs as functions of the input parameters, and in the inverse direction (predicting input parameters from the model outputs), to analyse the behaviour of a dynamic model of the mammalian circadian clock. Our results display a more complete cartography of how variation in input parameters is reflected in the temporal behaviour of multiple model outputs than has been previously reported. CONCLUSIONS: Our results indicated that the N-way HC-PLSR metamodelling provides a gain in insight into which parameters that are related to a specific model output behaviour, as well as variations in the model sensitivity to certain input parameters across the model output space. Moreover, the N-way approach allows a more transparent and detailed exploration of the temporal dimension of complex dynamic models, compared to alternative 2-way methods.

A Partial Least Squares based algorithm for parsimonious variable selection.

BACKGROUND: In genomics, a commonly encountered problem is to extract a subset of variables out of a large set of explanatory variables associated with one or several quantitative or qualitative response variables. An example is to identify associations between codon-usage and phylogeny based definitions of taxonomic groups at different taxonomic levels. Maximum understandability with the smallest number of selected variables, consistency of the selected variables, as well as variation of model performance on test data, are issues to be addressed for such problems. RESULTS: We present an algorithm balancing the parsimony and the predictive performance of a model. The algorithm is based on variable selection using reduced-rank Partial Least Squares with a regularized elimination. Allowing a marginal decrease in model performance results in a substantial decrease in the number of selected variables. This significantly improves the understandability of the model. Within the approach we have tested and compared three different criteria commonly used in the Partial Least Square modeling paradigm for variable selection; loading weights, regression coefficients and variable importance on projections. The algorithm is applied to a problem of identifying codon variations discriminating different bacterial taxa, which is of particular interest in classifying metagenomics samples. The results are compared with a classical forward selection algorithm, the much used Lasso algorithm as well as Soft-threshold Partial Least Squares variable selection. CONCLUSIONS: A regularized elimination algorithm based on Partial Least Squares produces results that increase understandability and consistency and reduces the classification error on test data compared to standard approaches.

Mining for genotype-phenotype relations in Saccharomyces using partial least squares.

BACKGROUND: Multivariate approaches are important due to their versatility and applications in many fields as it provides decisive advantages over univariate analysis in many ways. Genome wide association studies are rapidly emerging, but approaches in hand pay less attention to multivariate relation between genotype and phenotype. We introduce a methodology based on a BLAST approach for extracting information from genomic sequences and Soft- Thresholding Partial Least Squares (ST-PLS) for mapping genotype-phenotype relations. RESULTS: Applying this methodology to an extensive data set for the model yeast Saccharomyces cerevisiae, we found that the relationship between genotype-phenotype involves surprisingly few genes in the sense that an overwhelmingly large fraction of the phenotypic variation can be explained by variation in less than 1% of the full gene reference set containing 5791 genes. These phenotype influencing genes were evolving 20% faster than non-influential genes and were unevenly distributed over cellular functions, with strong enrichments in functions such as cellular respiration and transposition. These genes were also enriched with known paralogs, stop codon variations and copy number variations, suggesting that such molecular adjustments have had a disproportionate influence on Saccharomyces yeasts recent adaptation to environmental changes in its ecological niche. CONCLUSIONS: BLAST and PLS based multivariate approach derived results that adhere to the known yeast phylogeny and gene ontology and thus verify that the methodology extracts a set of fast evolving genes that capture the phylogeny of the yeast strains. The approach is worth pursuing, and future investigations should be made to improve the computations of genotype signals as well as variable selection procedure within the PLS framework.

Hierarchical cluster-based partial least squares regression (HC-PLSR) is an efficient tool for metamodelling of nonlinear dynamic models.

BACKGROUND: Deterministic dynamic models of complex biological systems contain a large number of parameters and state variables, related through nonlinear differential equations with various types of feedback. A metamodel of such a dynamic model is a statistical approximation model that maps variation in parameters and initial conditions (inputs) to variation in features of the trajectories of the state variables (outputs) throughout the entire biologically relevant input space. A sufficiently accurate mapping can be exploited both instrumentally and epistemically. Multivariate regression methodology is a commonly used approach for emulating dynamic models. However, when the input-output relations are highly nonlinear or non-monotone, a standard linear regression approach is prone to give suboptimal results. We therefore hypothesised that a more accurate mapping can be obtained by locally linear or locally polynomial regression. We present here a new method for local regression modelling, Hierarchical Cluster-based PLS regression (HC-PLSR), where fuzzy C-means clustering is used to separate the data set into parts according to the structure of the response surface. We compare the metamodelling performance of HC-PLSR with polynomial partial least squares regression (PLSR) and ordinary least squares (OLS) regression on various systems: six different gene regulatory network models with various types of feedback, a deterministic mathematical model of the mammalian circadian clock and a model of the mouse ventricular myocyte function. RESULTS: Our results indicate that multivariate regression is well suited for emulating dynamic models in systems biology. The hierarchical approach turned out to be superior to both polynomial PLSR and OLS regression in all three test cases. The advantage, in terms of explained variance and prediction accuracy, was largest in systems with highly nonlinear functional relationships and in systems with positive feedback loops. CONCLUSIONS: HC-PLSR is a promising approach for metamodelling in systems biology, especially for highly nonlinear or non-monotone parameter to phenotype maps. The algorithm can be flexibly adjusted to suit the complexity of the dynamic model behaviour, inviting automation in the metamodelling of complex systems.

Perceptions of German prison officials while visiting American correctional facilities: an example of a fruitful international collaboration.

In June 2007, a group of 15 prison psychologists, social workers, wardens, and correctional administrators from across Germany visited the northeast United States for the purpose of conducting tours of various kinds of correctional facilities. The trip was organized through a collaborative effort from the authors over a period of 2 years. This article describes the correctional facilities visited and observations made by the Germans during their correctional facility tours and will focus on the similarities and differences between the German and American correctional systems. The article clearly reflects how international academic collaborations can provide a variety of benefits for those willing to physically venture beyond boarders.

Predicting the fatty acid composition of milk: a comparison of two Fourier transform infrared sampling techniques.

In the present study a novel approach for Fourier transform infrared (FT-IR) characterization of the fatty acid composition of milk based on dried film measurements has been presented and compared to a standard FT-IR approach based on liquid milk measurements. Two hundred and sixty-two (262) milk samples were obtained from a feeding experiment, and the samples were measured with FT-IR as dried films as well as liquid samples. Calibrations against the most abundant fatty acids, CLA (i.e., 18:2cis-9, trans-11), 18:3cis-9, cis-12, cis-15, and summed fatty acid parameters were obtained for both approaches. The estimation errors obtained in the dried film calibrations were overall lower than the corresponding liquid sample calibrations. Similar and good calibrations (i.e., R(2) ranges from 0.82 to 0.94 (liquid samples) and from 0.88 to 0.97 (dried films)) for short-chain fatty acids (6:0-14:0), 18:1cis-9, SAT, MUFA, and iodine value were obtained by both approaches. However, the dried film approach was the only approach for which feasible calibrations (i.e., R(2) ranges from 0.78 to 0.93) were obtained for the major saturated fatty acids 16:0 and 18:0, the minor fatty acid features 4:0, CLA (i.e., 18:2cis-9, trans-11), PUFA, and the summed 18:1 trans isomers. For the dried film approach, logical spectral features were found to dominate the respective fatty acid calibration models. The preconcentration step of the dried film approach could be expected to account for a major part of the prediction improvements going from predictions in liquid milk to predictions in dried films. The dried film approach has a significant potential for use in high-throughput applications in industrial environments and might also serve as a valuable supplement for determination of genetic and breeding factors within research communities.

A high-throughput microcultivation protocol for FTIR spectroscopic characterization and identification of fungi.

Characterization and identification of fungi in food industry is an important issue both for routine analysis and trouble-shooting incidences. Present microbial techniques for fungal characterization suffer from a low throughput and are time consuming. In this study we present a protocol for high-throughput microcultivation and spectral characterization of fungi by Fourier transform infrared spectroscopy. For the study 11 species of in total five different fungal genera (Alternaria, Aspergillus, Mucor, Paecilomyces, and Phoma) were analyzed by FTIR spectroscopy. All the strains were isolated from trouble-shooting incidents in the production of low and high acid beverages. The cultivation was performed in malt extract broth (liquid medium) in a Bioscreen C system, allowing high-throughput cultivation of 200 samples at the same time. Mycelium was subsequently investigated by high-throughput Fourier transform infrared spectroscopy. Four spectral regions, fatty acids + lipid (3200-2800 cm(-1), 1300-1000 cm(-1)), protein-lipid (1800-1200 cm(-1)), carbohydrates (1200-700 cm(-1)) and "finger print" (900-700 cm(-1)) were evaluated for reproducibility and discrimination ability. The results show that all spectral regions evaluated can be used as spectroscopic biomarkers for differentiation of fungi by FTIR. The influence of different growth times on the ability of species discrimination by FTIR spectroscopy was investigated, and an optimal separation of all five genera was observed after five days of growth. This work presents a novel concept for high-throughput cultivation of fungi for FTIR spectroscopy that enables characterization or identification of hundreds of strains per day.

RMieS-EMSC correction for infrared spectra of biological cells: extension using full Mie theory and GPU computing.

In the field of biomedical infrared spectroscopy it is often desirable to obtain spectra at the cellular level. Samples consisting of isolated single biological cells are particularly unsuited to such analysis since cells are strong scatterers of infrared radiation. Thus measured spectra consist of an absorption component often highly distorted by scattering effects. It is now known that the predominant contribution to the scattering is Resonant Mie Scattering (RMieS) and recently we have shown that this can be corrected for, using an iterative algorithm based on Extended Multiplicative Signal Correction (EMSC) and a Mie approximation formula. Here we present an iterative algorithm that applies full Mie scattering theory. In order to avoid noise accumulation in the iterative algorithm a curve-fitting step is implemented on the new reference spectrum. The new algorithm increases the computational time when run on an equivalent processor. Therefore parallel processing by a Graphics Processing Unit (GPU) was employed to reduce computation time. The optimised RMieS-EMSC algorithm is applied to an IR spectroscopy data set of cultured single isolated prostate cancer (PC-3) cells, where it is shown that spectral distortions from RMieS are removed.