Procrustes cross-validation of multivariate regression models.

A generalization of Procrustes Cross-Validation - recently introduced novel approach for validation of chemometric models - is proposed. The generalized approach is faster than its predecessor by several orders of magnitude and can be used for validation of a wider range of models. Furthermore, it provides new tools for exploring the heterogeneity of the dataset, quality of cross-validation splits, presence of outliers, etc. The paper describes methodological aspects of the generalized approach, based on using Procrustean rules, the mathematical background, as well as presents practical results obtained using real chemical datasets.

Full-scale activated sludge transplantation reveals a highly resilient community structure.

Well-functioning and stable microbial communities are critical for the operation of activated sludge (AS) wastewater treatment plants (WWTPs). Bioaugmentation represents a potentially useful approach to recover deteriorated systems or to support specific AS processes, but its application in full-scale WWTPs is generally problematic. We conducted a massive transplantation (in one day) exchanging AS from a donor to a recipient full-scale WWTP with similar process type (biological removal of nitrogen and phosphorus) and performance, but with differences in microbial community structure. The treatment performance in the recipient plant was not compromised and the effluent quality remained stable. The AS community structure of the recipient plant was initially very similar to the donor AS, but it almost completely restored the pre-transplantation structure approximately 40 days after transplantation, corresponding to 3 times the solid retention time. Most of the unique species of donor AS added to recipient AS disappeared quickly, although some disappeared more slowly the following months, indicating some survival and potentially a time limited function in the recipient plant. Moreover, the addition in higher abundance of most species already present in the recipient AS (e.g., the polyphosphate accumulating organisms) or the reduction of the abundance of unwanted bacteria (e.g., filamentous bacteria) in the recipient plant was not successful. Moreover, we observed similar abundance patterns after transplantation for species belonging to different functional guilds, so we did not observe an increase of the functional redundancy. Investigations of the microbial community structure in influent wastewater revealed that for some species the abundance trends in the recipient plant were closely correlated to their abundance in the influent. We showed that a very resilient microbial community was responsible for the outcome of the transplantation of AS at full-scale WWTP, potentially as a consequence of mass-immigration from influent wastewater. The overall results imply that massive transplantation of AS across different WWTPs is not a promising strategy to permanently solve operational problems. However, by choosing a compatible AS donor, short term mitigation of serious operational problems may be possible.

Data reduction by randomization subsampling for the study of large hyperspectral datasets.

Large amount of information in hyperspectral images (HSI) generally makes their analysis (e.g., principal component analysis, PCA) time consuming and often requires a lot of random access memory (RAM) and high computing power. This is particularly problematic for analysis of large images, containing millions of pixels, which can be created by augmenting series of single images (e.g., in time series analysis). This tutorial explores how data reduction can be used to analyze time series hyperspectral images much faster without losing crucial analytical information. Two of the most common data reduction methods have been chosen from the recent research. The first one uses a simple randomization method called randomized sub-sampling PCA (RSPCA). The second implies a more robust randomization method based on local-rank approximations (rPCA). This manuscript exposes the major benefits and drawbacks of both methods with the spirit of being as didactical as possible for a reader. A comprehensive comparison is made considering the amount of information retained by the PCA models at different compression degrees and the performance time. Extrapolation is also made to the case where the effect of time and any other factor are to be studied simultaneously.

Mass-immigration determines the assembly of activated sludge microbial communities.

The assembly of bacterial communities in wastewater treatment plants (WWTPs) is affected by immigration via wastewater streams, but the impact and extent of bacterial immigrants are still unknown. Here, we quantify the effect of immigration at the species level in 11 Danish full-scale activated sludge (AS) plants. All plants have different source communities but have very similar process design, defining the same overall environmental growth conditions. The AS community composition in each plant was strongly reflected by the corresponding influent wastewater (IWW) microbial composition. Most species in AS across the plants were detected and quantified in the corresponding IWW, allowing us to identify their fate in the AS: growing, disappearing, or surviving. Most of the abundant species in IWW disappeared in AS, so their presence in the AS biomass was only due to continuous mass-immigration. In AS, most of the abundant growing species were present in the IWW at very low abundances. We predicted the AS species abundances from their abundance in IWW by using a partial least square regression model. Some species in AS were predicted by their own abundance in IWW, while others by multiple species abundances. Detailed analyses of functional guilds revealed different prediction patterns for different species. We show, in contrast to the present understanding, that the AS microbial communities were strongly controlled by the IWW source community and could be quantitatively predicted by taking into account immigration. This highlights a need to revise the way we understand, design, and manage the microbial communities in WWTPs.

Characterizing the growing microorganisms at species level in 46 anaerobic digesters at Danish wastewater treatment plants: A six-year survey on microbial community structure and key drivers.

Anaerobic digestion (AD) is a key technology at many wastewater treatment plants (WWTPs) for converting primary and surplus activated sludge to methane-rich biogas. However, the limited number of surveys and the lack of comprehensive datasets have hindered a deeper understanding of the characteristics and associations between key variables and the microbial community composition. Here, we present a six-year survey of 46 anaerobic digesters, located at 22 WWTPs in Denmark, which is the first and largest known study of the microbial ecology of AD at WWTPs at a regional scale. For three types of AD (mesophilic, mesophilic with thermal hydrolysis pretreatment, and thermophilic), we present the typical value range of 12 key parameters including operational variables and performance parameters. High-resolution bacterial and archaeal community analyses were carried out at species level using amplicon sequencing of >1,000 samples and the new ecosystem-specific MiDAS 3 reference database. We detected 42 phyla, 1,600 genera, and 3,584 species in the bacterial community, where 70% of the genera and 93% of the species represented environmental taxa that were only classified based on MiDAS 3 de novo placeholder taxonomy. More than 40% of the bacterial species were found not to grow in the mesophilic and thermophilic digesters and were only present due to immigration with the feed sludge. Ammonium concentration was the main driver shaping the bacterial community while temperature and pH were main drivers for the archaea in the three types of ADs. Sub-setting for the growing microbes improved significantly the correlation analyses and revealed the main drivers for the presence of specific species. Within mesophilic digesters, feed sludge composition and other key parameters (organic loading rate, biogas yield, and ammonium concentration) correlated with specific growing species. This survey provides a comprehensive insight into community structure at species level, providing a foundation for future studies of the ecological significance/characteristics and function of the many novel or poorly described taxa.

Coupling Analytical Models and Machine Learning Methods for Fast and Reliable Resolution of Effects in Multifrequency Eddy-Current Sensors.

Eddy current (EC) measurements, widely used for diagnostics of conductive materials, are highly dependent on physical properties and geometry of a sample as well as on a design of an EC-sensor. For a sensor of a given design, the conductivity and thickness of a sample as well as the gap between the sample and the sensor (lift-off) are the most influencing parameters. Estimation of these parameters, based on signals acquired from the sensor, is quite complicated in case when all three parameters are unknown and may vary. In this paper, we propose a machine learning based approach for solving this problem. The approach makes it possible to avoid time and resource-consuming computations and does not require experimental data for training of the prediction models. The approach was tested using independent sets of measurements from both simulated and real experimental data.

Combining Laser-Induced Breakdown Spectroscopy (LIBS) and Visible Near-Infrared Spectroscopy (Vis-NIRS) for Soil Phosphorus Determination.

Conventional wet chemical methods for the determination of soil phosphorus (P) pools, relevant for environmental and agronomic purposes, are labor-intensive. Therefore, alternative techniques are needed, and a combination of the spectroscopic techniques-in this case, laser-induced breakdown spectroscopy (LIBS)-and visible near-infrared spectroscopy (vis-NIRS) could be relevant. We aimed at exploring LIBS, vis-NIRS and their combination for soil P estimation. We analyzed 147 Danish agricultural soils with LIBS and vis-NIRS. As reference measurements, we analyzed water-extractable P (Pwater), Olsen P (Polsen), oxalate-extractable P (Pox) and total P (TP) by conventional wet chemical protocols, as proxies for respectively leachable, plant-available, adsorbed inorganic P, and TP in soil. Partial least squares regression (PLSR) models combined with interval partial least squares (iPLS) and competitive adaptive reweighted sampling (CARS) variable selection methods were tested, and the relevant wavelengths for soil P determination were identified. LIBS exhibited better results compared to vis-NIRS for all P models, except for Pwater, for which results were comparable. Model performance for both the LIBS and vis-NIRS techniques as well as the combined LIBS-vis-NIR approach was significantly improved when variable selection was applied. CARS performed better than iPLS in almost all cases. Combined LIBS and vis-NIRS models with variable selection showed the best results for all four P pools, except for Pox where the results were comparable to using the LIBS model with CARS. Merging LIBS and vis-NIRS with variable selection showed potential for improving soil P determinations, but larger and independent validation datasets should be tested in future studies.

Procrustes Cross-Validation-A Bridge between Cross-Validation and Independent Validation Sets.

In this paper, we propose a new approach for validation of chemometric models. It is based on k-fold cross-validation algorithm, but in contrast to conventional cross-validation, our approach makes it possible to create a new dataset, which carries sampling uncertainty estimated by the cross-validation procedure. This dataset, called a pseudo-validation set, can be used similar to an independent test set, giving a possibility to compute residual distances, explained variance, scores, and other results, which cannot be obtained in the conventional cross-validation. The paper describes theoretical details of the proposed approach and its implementation as well as presents experimental results obtained using simulated and real chemical datasets.

Candidatus Amarolinea and Candidatus Microthrix Are Mainly Responsible for Filamentous Bulking in Danish Municipal Wastewater Treatment Plants.

Filamentous bulking is a common serious operational problem leading to deteriorated sludge settling that has long been observed in activated sludge biological wastewater treatment systems. A number of bacterial genera found therein possess filamentous morphology, where some have been shown to be implicated in bulking episodes (e.g., Ca. Microthrix), the impact of many others is still not clear. In this study we performed a survey of 17 Danish municipal wastewater treatment plants (WWTPs) with nutrient removal using 16S rRNA amplicon sequencing over a period of 13 years, where all known filamentous bacteria from 30 genera were analyzed. The filamentous community constituted on average 13 ± 6%, and up to 43% of total read abundance with the same genera common to all plants. Ca. Microthrix and several genera belonging to phylum Chloroflexi were among the most abundant filamentous bacteria. The effect of filamentous bacteria on sludge settling properties was analyzed using measurements of the diluted sludge volume index (DSVI). Strong positive correlations with DSVI were observed only for Ca. Microthrix and Ca. Amarolinea, the latter being a novel, recently characterized genus belonging to the phylum Chloroflexi. The bulking potential of other filamentous bacteria was not significant despite their presence in many plants. Low phylogenetic diversity was observed for both Ca. Microthrix and Ca. Amarolinea, making physiological characterization of individual species and potential development of control strategies more feasible. In this study we show that, despite the high diversity of filamentous phylotypes in Danish WWTPs, only few of them were responsible for severe bulking episodes.

Assessment of PCBs and exposure risk to infants in breast milk of primiparae and multiparae mothers in an electronic waste hot spot and non-hot spot areas in Ghana.

The aim of the study was to assess the levels of PCBs in the breast milk of some Ghanaian women at suspected hotspot and relatively non-hotspot areas and to find out if the levels of these PCBs pose any risk to the breastfed infants. A total of 128 individual human breast milk were sampled from both primiparae and multiparae mothers. The levels of PCBs in the milk samples were compared. Some of these mothers (105 individuals) work or reside in and around Agbogbloshie (hot-spot), the largest electric and electronic waste dump and recycling site in Accra, Ghana. Others (23 donor mothers) also reside in and around Kwabenya (non-hotspot) which is a mainly residential area without any industrial activities. Samples were analyzed using GC-MS/MS. The total mean levels and range of Σ7PCBs were 3.64ng/glipidwt and ˂LOD-29.20ng/glipidwt, respectively. Mean concentrations from Agbogbloshie (hot-spot area) and Kwabenya (non-hotspot areas) were 4.43ng/glipidwt and 0.03ng/glipidwt, respectively. PCB-28 contributed the highest of 29.5% of the total PCBs in the milk samples, and PCB-101 contributed the lowest of 1.74%. The estimated daily intake of PCBs and total PCBs concentrations in this work were found to be lower as compared to similar studies across the world. The estimated hazard quotient using Health Canada's guidelines threshold limit of 1µg/kgbw/day showed no potential health risk to babies. However, considering minimum tolerable value of 0.03µg/kgbw/day defined by the Agency for Toxic Substances and Disease Registry (ATSDR), the values of some mothers were found to be at the threshold limit. This may indicate a potential health risk to their babies. Mothers with values at the threshold levels of the minimum tolerable limits are those who work or reside in and around the Agbogbloshie e-waste site.

Metabotyping Patients' Journeys Reveals Early Predisposition to Lung Injury after Cardiac Surgery.

Cardiovascular disease is the leading cause of death worldwide and patients with severe symptoms undergo cardiac surgery. Even after uncomplicated surgeries, some patients experience postoperative complications such as lung injury. We hypothesized that the procedure elicits metabolic activity that can be related to the disease progression, which is commonly observed two-three days postoperatively. More than 700 blood samples were collected from 50 patients at nine time points pre-, intra-, and postoperatively. Dramatic metabolite shifts were observed during and immediately after the intervention. Prolonged surgical stress was linked to an augmented anaerobic environment. Time series analysis showed shifts in purine-, nicotinic acid-, tyrosine-, hyaluronic acid-, ketone-, fatty acid, and lipid metabolism. A characteristic 'metabolic biosignature' was identified correlating with the risk of developing postoperative complications two days before the first clinical signs of lung injury. Hence, this study demonstrates the link between intra- and postoperative time-dependent metabolite changes and later postoperative outcome. In addition, the results indicate that metabotyping patients' journeys early, during or just after the end of surgery, may have potential impact in hospitals for the early diagnosis of postoperative lung injury, and for the monitoring of therapeutics targeting disease progression.

Classification of maize kernels using NIR hyperspectral imaging.

NIR hyperspectral imaging was evaluated to classify maize kernels of three hardness categories: hard, medium and soft. Two approaches, pixel-wise and object-wise, were investigated to group kernels according to hardness. The pixel-wise classification assigned a class to every pixel from individual kernels and did not give acceptable results because of high misclassification. However by using a predefined threshold and classifying entire kernels based on the number of correctly predicted pixels, improved results were achieved (sensitivity and specificity of 0.75 and 0.97). Object-wise classification was performed using two methods for feature extraction - score histograms and mean spectra. The model based on score histograms performed better for hard kernel classification (sensitivity and specificity of 0.93 and 0.97), while that of mean spectra gave better results for medium kernels (sensitivity and specificity of 0.95 and 0.93). Both feature extraction methods can be recommended for classification of maize kernels on production scale.

Monitoring of whey quality with NIR spectroscopy--a feasibility study.

The possibility of using near-infrared (NIR) spectroscopy for monitoring of liquid whey quality parameters during protein production process has been tested. The parameters included total solids, lactose, protein and fat content. The samples for the experiment were taken from real industrial processes and had a large variability for most of the parameters. Partial Least Squares (PLS) regression was used to make the prediction models based on NIR spectra taken at 30 and 40°C. Using proper wavelength range allowed to get models for prediction of fat, protein and amount of total solids with very high precision and accuracy. The lactose was found to be the most challenging parameter.

Determination of fat and total protein content in milk using conventional digital imaging.

The applicability of conventional digital imaging to quantitative determination of fat and total protein in cow's milk, based on the phenomenon of light scatter, has been proved. A new algorithm for extracting features from digital images of milk samples has been developed. The algorithm takes into account spatial distribution of light, diffusely transmitted through a sample. The proposed method has been tested on two sample sets prepared from industrial raw milk standards, with variable fat and protein content. Partial Least-Squares (PLS) regression on the features calculated from images of monochromatically illuminated milk samples resulted in models with high prediction performance when analysed the sets separately (best models with cross-validated R(2)=0.974 for protein and R(2)=0.973 for fat content). However when analysed the sets jointly with the obtained results were significantly worse (best models with cross-validated R(2)=0.890 for fat content and R(2)=0.720 for protein content). The results have been compared with previously published Vis/SW-NIR spectroscopic study of similar samples.

Predicting apple (cv. Elshof) postharvest dry matter and soluble solids content with near infrared spectroscopy.

BACKGROUND: Fruit dry matter (DM) and soluble solids content (SSC) are primarily composed of carbohydrate and are standard parameters for assessing quality. Near infrared spectroscopy provides potential for non-destructive fruit quality analysis but the collinearity between DM and SSC is an issue for prediction. Shorter wavelength spectra have been used for the prediction of fruit DM and SSC, but radiation between 1000 and 2500 nm may be suitable for distinguishing between the two forms of carbohydrate. RESULTS: Spectra and DM and SSC samples were taken for a total of 450 'Elshof' apples 30, 58 and 93 days after harvest. Regression models were built using the interval partial least squares method. Prediction models for DM and SSC for each day yielded R² values between 0.63 and 0.86 and residual predictive deviations (RPDs) between 1.7 and 2.7 for DM, and R² = 0.76-0.85 and RPDs = 2.2-2.6 for SSC. CONCLUSION: Model RPD values were not high enough for general quantitative predictions, although they compare well to previous work. Certain factors affected model success, including changes in fruit physiology over time and the range of reference data. The complexity of absorbance spectra for DM and SSC plus their strong correlation suggests that prediction models cannot easily distinguish between soluble and non-soluble forms of carbohydrate.