Feasibility study on the potential of visible and near infrared reflectance spectroscopy to measure alpaca fibre characteristics

Visible (Vis) and near infrared (NIR) reflectance spectroscopy is a rapid and non-destructive technique that has found many applications in assessing the quality of agricultural commodities, including wool. In this study, Vis and NIR spectroscopy combined with multivariate data analysis was investigated regarding its feasibility in predicting a range of fibre characteristics in raw alpaca wool samples. Mid-side samples (n = 149) were taken from alpacas from a range of colours and ages at shearing time over 4 years (2000 to 2004) and subsequently analysed for fibre characteristics such as mean fibre diameter (MFD) and standard deviation (and coefficient of variation), spin fineness, curvature degree (and standard deviation), comfort factor, medullation percentage (by weight and number in white samples only) using traditional reference laboratory testing methods. Samples were scanned in a large cuvette using a FOSS NIRSystems 6500 monochromator instrument in reflectance mode in the Vis and NIR regions (400 to 2500 nm). Partial least squares (PLS) regression was used to develop a number of calibration models between the spectral and reference data. Mathematical pre-treatment of the spectra (second derivative) as well as various combinations of wavelength range were used in model development. The best calibration model was found when using the NIR region (1100 to 2500 nm) for the prediction of MFD, which had a coefficient of determination in cross-validation (R2) of 0.88 with a root mean square standard error of cross validation (RMSECV) of 2.62 μm. The results show the NIR technique to have promise as a semi-quantitative method for screening purposes. The lack of grease in alpaca wool samples suggests that the technique might find ready application as a rapid measurement technique for preliminary classing of shorn fleeces or, if used directly on the animal, the technology might offer an objective tool to assist in the selection of animals in breeding programmes or shows.     

近红外光谱技术预测猪肉的pH、嫩度和蒸煮损失

以冷却猪肉为研究对象,评价近红外光谱（NIR）技术用于肉类物理特性预测的可行性以及不同的光谱处理方法和建模方法对预测准确性的影响。试样取自排酸24h的同一批猪胴体的小里脊肉,采集4000—10000cm-1的光谱。经外部验证的偏最小二乘（PLS）模型在预测pH时表现出良好的相关性（Rc^2=0．88,Rp^2=0．80,SEC=0．08,SEP=0．084）,嫩度与蒸煮损失模型的相关性分别是Rc^2=0．50和0．57,R;=0．34和0．50。在各种光谱预处理方法中,平滑处理结合多元散射校正（MSC）或标准正态变量变换（SNV）的效果最好。     

Quantitation of drug content in a low dosage formulation by transmission near infrared spectroscopy

A transmission near infrared (NIR) spectroscopic method has been developed for the nondestructive determination of drug content in tablets with less than 1% weight of active ingredient per weight of formulation (m/m) drug content. Tablets were manufactured with drug concentrations of ∼0.5%, 0.7%, and 1.0% (m/m) and ranging in drug content from 0.71 to 2.51 mg per tablet. Transmission NIR spectra were obtained for 110 tablets that constituted the training set for the calibration model developed with partial least squares regression. The reference method for the calibration model was a validated UV spectrophotometric method. Several data preprocessing methods were used to reduce the effect of scattering on the NIR spectra and base the calibration model on spectral changes related to the drug concentration changes. The final calibration model included the spectral range from 11 216 to 8662 cm⁻¹ the standard normal variate (SNV), and first derivative spectral pretreatments. This model was used to predict an independent set of 48 tablets with a root mean standard error of prediction (RMSEP) of 0.14 mg, and a bias of only −0.05 mg per tablet. The study showed that transmission NIR spectroscopy is a viable alternative for nondestructive testing of low drug content tablets, available for the analysis of large numbers of tablets during process development and as a tool to detect drug agglomeration and evaluate process improvement efforts. Published: March 24, 2006     

The prediction of intake potential and organic matter digestibility of grass silages by near infrared spectroscopy analysis of undried samples

A study was undertaken to examine a range of sample preparation and near infrared reflectance spectroscopy (NIRS) methodologies, using undried samples, for predicting organic matter digestibility (OMD g kg⁻¹) and ad libitum intake (g kg⁻¹ W^0.75) of grass silages. A total of eight sample preparation/NIRS scanning methods were examined involving three extents of silage comminution, two liquid extracts and scanning via either external probe (1100-2200 nm) or internal cell (1100-2500 nm). The spectral data (log 1/R) for each of eight methods were examined by three regression techniques each with a range of data transformations. The 136 silages used in the study were obtained from farms across Northern Ireland, over a two year period, and had in vivo OMD (sheep) and ad libitum intake (cattle) determined under uniform conditions. In the comparisons of the eight sample preparation/scanning methods, and the differing mathematical treatments of the spectral data, the sample population was divided into calibration (n = 91) and validation (n = 45) sets. The standard error of performance (SEP) on the validation set was used in comparisons of prediction accuracy. Across all 8 sample preparation/scanning methods, the modified partial least squares (MPLS) technique, generally minimized SEP's for both OMD and intake. The accuracy of prediction also increased with degree of comminution of the forage and with scanning by internal cell rather than external probe. The system providing the lowest SEP used the MPLS regression technique on spectra from the finely milled material scanned through the internal cell. This resulted in SEP and R² (variance accounted for in validation set) values of 24 (g/kg OM) and 0.88 (OMD) and 5.37 (g/kg W^0.75) and 0.77 (intake) respectively. These data indicate that with appropriate techniques NIRS scanning of undried samples of grass silage can produce predictions of intake and digestibility with accuracies similar to those achieved previously using NIRS with dried samples.     

Use of near infrared spectroscopy in online-monitoring of feeding substrate quality in anaerobic digestion

In order to keep the anaerobic process stably and uniformly producing biogas it needs to be supplied with either an even amount of substrate of stable quality or varying amounts according to variations in quality. Feeding amounts are usually adjusted manually as a reaction to changing rates of biogas production. Continuous information about the actual substrate quality is not available and feedstuff analyses are costly. Aim of this study was to assess the feasibility of near infrared spectroscopic (NIRS) online monitoring of substrate quality in order to find ways towards more exact control of biogas plant feeding. A NIRS sensor system was designed, constructed and calibrated for continuous monitoring of (RMSECV in brackets) dry matter (DM) (0.75 %fresh matter (FM)), volatile solids (0.74 %FM), crude fat (0.09 %FM), crude protein (0.22 %FM), crude fiber (1.50 %DM) and nitrogen-free extracts (0.93 %FM) of maize silage.     

Rapid monitoring of recombinant antibody production by mammalian cell cultures using fourier transform infrared spectroscopy and chemometrics

Fourier transform infrared (FT‐IR) spectroscopy combined with multivariate statistical analyses was investigated as a physicochemical tool for monitoring secreted recombinant antibody production in cultures of Chinese hamster ovary (CHO) and murine myeloma non‐secreting 0 (NS0) cell lines. Medium samples were taken during culture of CHO and NS0 cells lines, which included both antibody‐producing and non‐producing cell lines, and analyzed by FT‐IR spectroscopy. Principal components analysis (PCA) alone, and combined with discriminant function analysis (PC‐DFA), were applied to normalized FT‐IR spectroscopy datasets and showed a linear trend with respect to recombinant protein production. Loadings plots of the most significant spectral components showed a decrease in the C–O stretch from polysaccharides and an increase in the amide I band during culture, respectively, indicating a decrease in sugar concentration and an increase in protein concentration in the medium. Partial least squares regression (PLSR) analysis was used to predict antibody titers, and these regression models were able to predict antibody titers accurately with low error when compared to ELISA data. PLSR was also able to predict glucose and lactate amounts in the medium samples accurately. This work demonstrates that FT‐IR spectroscopy has great potential as a tool for monitoring cell cultures for recombinant protein production and offers a starting point for the application of spectroscopic techniques for the on‐line measurement of antibody production in industrial scale bioreactors. Biotechnol. Bioeng. 2010; 106: 432–442. © 2010 Wiley Periodicals, Inc.     

Characterization of mammalian cell culture raw materials by combining spectroscopy and chemometrics

Two of the primary issues with characterizing the variability of raw materials used in mammalian cell culture, such as wheat hydrolysate, is that the analyses of these materials can be time consuming, and the results of the analyses are not straightforward to interpret. To solve these issues, spectroscopy can be combined with chemometrics to provide a quick, robust and easy to understand methodology for the characterization of raw materials; which will improve cell culture performance by providing an assessment of the impact that a given raw material will have on final product quality. In this study, four spectroscopic technologies: near infrared spectroscopy, middle infrared spectroscopy, Raman spectroscopy, and fluorescence spectroscopy were used in conjunction with principal component analysis to characterize the variability of wheat hydrolysates, and to provide evidence that the classification of good and bad lots of raw material is possible. Then, the same spectroscopic platforms are combined with partial least squares regressions to quantitatively predict two cell culture critical quality attributes (CQA): integrated viable cell density and IgG titer. The results showed that near infrared (NIR) spectroscopy and fluorescence spectroscopy are capable of characterizing the wheat hydrolysate's chemical structure, with NIR performing slightly better; and that they can be used to estimate the raw materials’ impact on the CQAs. These results were justified by demonstrating that of all the components present in the wheat hydrolysates, six amino acids: arginine, glycine, phenylalanine, tyrosine, isoleucine and threonine; and five trace elements: copper, phosphorus, molybdenum, arsenic and aluminum, had a large, statistically significant effect on the CQAs, and that NIR and fluorescence spectroscopy performed the best for characterizing the important amino acids. It was also found that the trace elements of interest were not characterized well by any of the spectral technologies used; however, the trace elements were also shown to have a less significant effect on the CQAs than the amino acids. © 2017 The Authors Biotechnology Progress published by Wiley Periodicals, Inc. on behalf of American Institute of Chemical Engineers, 33:1127–1138, 2017     

应用近红外光谱法估测小麦叶片糖氮比

糖氮比能够反映作物碳氮代谢的协调程度,及时、准确地监测糖氮比对于作物氮素营养诊断和调控具有重要意义.本研究以不同年份、品种、施氮水平的小麦大田试验为基础,获取鲜叶和粉末状干叶近红外(NIR)光谱及糖氮比信息,分别运用偏最小二乘法(partial least squares, PLS)、BP神经网络(back propagation neural network, BPNN)和小波神经网络(wavelet neural network, WNN)3种方法建立了小麦叶片糖氮比预测模型,并利用随机选择的样品集对所建模型进行测试和检验.结果表明： 小麦鲜叶光谱模型预测性能不佳;而干叶片预测模型表现了较好的准确性,在1655~2378 nm谱区范围内基于3种方法构建的干叶粉末糖氮比估算模型,其预测均方根误差均低于0.3%,决定系数均高于0.9.比较而言,WNN法表现最佳.总体显示,近红外光谱法可以准确预测小麦叶片糖氮比状况,为科学诊断糖氮比提供了理论基础和技术途径.     

红外光谱技术在滇龙胆栽培方式选择上的应用

该研究采用傅里叶变换红外光谱结合化学计量学,对条播、撒播、剪根后移栽、扦插和剪枝后移栽的滇龙胆进行了分析,以筛选滇龙胆的最佳栽培方式。结果表明:(1)不同栽培方式的滇龙胆原始谱图在峰形、峰位和峰强上有一定差异;用小波去噪法对光谱进行优化处理并进行偏最小二乘判别分析(Partial least squares discriminant analysis,PLS-DA),能较好地区分不同栽培方式的滇龙胆样品,PLS-DA二维得分图显示同一栽培方式的样品聚在一起,表明相同栽培方式的滇龙胆化学组成和含量差异较小;播种滇龙胆样品(条播和撒播)距离较近,移栽滇龙胆样品(剪根、扦插和剪枝)距离较近,而播种和移栽滇龙胆样品距离较远,表明栽培方式对滇龙胆化学成分的积累有影响。(2)滇龙胆四种主要成分总含量大小依次是剪枝剪根撒播条播扦插,除剪根后移栽,剪枝后移栽滇龙胆中四种主要成分总含量显著高于其他栽培方式下的滇龙胆(P0.05),剪枝后移栽滇龙胆质量最佳。(3)以液相数据为参考值,采用正交信号校正—偏最小二乘回归模型预测不同栽培模式滇龙胆中龙胆苦苷、马钱苷酸、獐牙菜苦苷和当药苷的含量。校正集和验证集的决定系数(R2)均大于0.90,校正均方根误差、交叉验证均方差和预测均方根误差均小于1.65,模型相关性和预测效果好,该方法对红外光谱分析在中药领域的推广应用提供了参考。     

Online prediction of fatty acid profiles in crossbred Limousin and Aberdeen Angus beef cattle using near infrared reflectance spectroscopy

The objective of this study was to examine the online use of near infrared reflectance (NIR) spectroscopy to estimate the concentration of individual and groups of fatty acids (FA) as well as intramuscular fat (IMF) in crossbred Aberdeen Angus (AA×) and Limousin (LIM×) cattle. This was achieved by direct application of a fibre-optic probe to the muscle immediately after exposing the meat surface in the abattoir at 48 h post mortem. Samples of M. longissimus thoracis from 88 AA× and 106 LIM× were scanned over the NIR spectral range from 350 to 1800 nm and samples of the M. longissimus lumborum were analysed for IMF content and FA composition. Statistically significant differences (P < 0.001) were observed in most FA between the two breeds studied, with FA concentration being higher in AA× meat mainly. NIR calibrations, tested by cross-validation, showed moderate to high predictability in LIM× meat samples for C16:0, C16:1, C18:0, trans11 C18:1, C18:1, C18:2 n-6, C20:1, cis9, trans11 C18:2, SFA (saturated FA), MUFA (monounsaturated FA), PUFA (polyunsaturated FA) and IMF content with R(2) (SE(CV), mg/100 g muscle) of 0.69 (146), 0.69 (28), 0.71 (62), 0.70 (8.1), 0.76 (192), 0.65 (13), 0.71 (0.9), 0.71 (2.9), 0.68 (235), 0.75 (240), 0.64 (17) and 0.75 (477), respectively. FA such as C14:0, C18:3 n-3, C20:4 n-6, C20:5 n-3, C22:6 n-3, n-6 and n-3 were more difficult to predict by NIR in these LIM× samples (R(2) = 0.12 to 0.62; SECV = 0.5 to 26 mg/100 g muscle). In contrast, NIR showed low predictability for FA in AA× beef samples. In particular for LIM×, the correlations of NIR measurements and several FA in the range from 0.81 to 0.87 indicated that the NIR spectroscopy is a useful online technique for the early, fast and relatively inexpensive estimation of FA composition in the abattoir.     

NucleoCounter—An efficient technique for the determination of cell number and viability in animal cell culture processes

The NucleoCounter is a novel, portable cell counting device based on the principle of fluorescence microscopy. The present work establishes its use with animal cells and checks its reliability, consistency and accuracy in comparison with other cytometric techniques. The main advantages of this technique are its ability to handle a large number of samples with a high degree of precision and its simplicity and specificity in detecting viable cells quantitatively in a heterogeneous culture. The work addresses and overcomes the problems of subjectivity, and some of the inherent sampling errors associated with using the traditional haemocytometer and Trypan Blue exclusion method. NucleoCounter offers reduced intra- and inter-observer variation as well as consistency in repetitive analysis that establishes it as an efficient and highly potential device for at-line monitoring of animal cell processes. Furthermore, since the only manual steps required are sample aspiration and mixing with two reagents, it is feasible that the whole method could be automated and brought on-line for process monitoring and control.     

Biochemical impact of solar radiation exposure on human keratinocytes monitored by Raman spectroscopy; effects of cell culture environment

Understanding and amelioration of the effects of solar radiation exposure are critical in preventing the occurrence of skin cancer. Towards this end, many studies have been conducted in 2D cell culture models under simplified and unrealistic conditions. 3D culture models better capture the complexity of in vivo physiology, although the effects of the 3D extracellular matrix have not been well studied. Monitoring the instantaneous and resultant cellular responses to exposure, and the influence of the 3D environment, could provide an enhanced understanding of the fundamental processes of photocarcinogenesis. This work presents an analysis of the biochemical impacts of simulated solar radiation (SSR) occurring in immortalised human epithelial keratinocytes (HaCaT), in a 3D skin model, compared to 2D culture. Cell viability was monitored using the Alamar Blue colorimetric assay (AB), and the impact of the radiation exposure, at the level of the biomolecular constituents (nucleic acids and proteins), were evaluated through the combination of Raman microspectroscopy and multivariate statistical analysis. The results suggest that SSR exposure induces alterations of the conformational structure of DNA as an immediate impact, whereas changes in the protein signature are primarily seen as a subsequent response.     

Whole body and tissue imaging of the butyrylcholinesterase knockout mouse injected with near infrared dye labeled butyrylcholinesterase

Butyrylcholinesterase (BChE) has proven to be an effective bioscavenger against nerve agents and organophosphates. Phase I safety trials of human BChE are currently being conducted and large-scale production of recombinant BChE is underway. Information on the real-time distribution of BChE from the injection site has not been well characterized. This study utilized the BChE nullizygote (BChE-/-) mouse and tetrameric equine BChE labeled with LI-COR((R)) fluorescent IRDye 800CW to track, quantify and determine the retention time of BChE in vivo following intramuscular injection. In vivo images were acquired with Xenogen's IVIS((R)) 200 imager and the LI-COR Odyssey((R)) Imaging System fitted with the MousePODtrade mark. Plasma and tissues were tested for BChE activity. The 2mg of BChE spread from the injection site to heart, liver, intestine, kidneys, lungs, salivary glands, and muscle, but did not enter the brain or the skin. Fluorescence intensity in organs and BChE activity in plasma peaked on day 1. BChE activity in plasma was undetectable by day 16, at a time when there was still significant fluorescent signal and BChE activity in the liver (0.32units/g), injected quadriceps (0.13units/g) and in most of the organs analyzed. It is concluded that the tetrameric BChE glycoprotein of 340kDa diffuses from the muscle injection site to blood and peripheral organs and has a longer residence time in the organs than in blood.     

An automated approach for fringe frequency estimation and removal in infrared spectroscopy and hyperspectral imaging of biological samples

In infrared spectroscopy of thin film samples, interference introduces distortions in spectra, commonly referred to as fringes. Fringes may alter absorbance peak ratios, which hampers the spectral analysis. We have previously introduced extended multiplicative signal correction (EMSC) for fringes correction. In the current article, we provide a robust open-source algorithm for fringe correction in infrared spectroscopy and propose several improvements to the Fringe EMSC model. The suggested algorithm achieves a more precise fringe frequency estimation by mean centering of the measured spectrum and applying a window function prior to the Fourier transform. It selects two frequencies from a user defined number of maxima in the Fourier domain. The improved Fringe EMSC algorithm is validated on two experimental datasets, one of them being a hyperspectral image. Techniques for separating sample spectra from background spectra in hyperspectral images, and techniques to identify spectra affected by fringes are also provided.     

Evaluation of the bony repair in rat cranial defect using near infrared reflectance spectroscopy and discriminant analysis

We set out to determine whether near infrared reflectance spectroscopy (NIRS) combined with principal component analysis–linear discriminant analysis (LDA) or, variable selection techniques employing successive projection algorithm or genetic algorithm (GA) could evaluate the bone repair in cranial critical‐size (5 mm) defect after stimulation with collagen sponge scaffold and/or infrared low‐level laser therapy directly on the local. Forty‐five Winstar rats were divided into nine groups of five each, namely: group H – healthy, n = 5 (without treatment and without cranial critical‐size defect), (GI positive control – n = 5, 21 days or n = 5, 30 days) without treatment and with cranial critical‐size defect; (GII‐n = 5, 21 days or n = 5, 30 days) cranial critical‐size defect filled with collagen sponge scaffold; (GIII–n = 5, 21 days or n = 5, 30 days) cranial critical‐size defect submitted to low‐level laser therapy; (GIV–n = 5, 21 days or n = 5, 30 days) cranial critical‐size defect submitted to combined collagen sponge scaffold + low‐level laser therapy treatment. In relation to the histological analysis, the collagen sponge scaffold + low‐level laser therapy treatment group (GIV) 30 days showed the best result with the presence of secondary bone, immature bone (osteoid) and newly formed connective tissue (periosteum). GA–LDA model also successfully classified control class of the others classes. Thus, the results provided by the good‐quality classification model revealed the feasibility of NIRS for application to evaluation of the wound healing in rat cranial defect, thanks to the short analysis time of a few seconds and nondestructive advantages of NIRS as an alternative approach for bone repair purposes. © 2017 American Institute of Chemical Engineers Biotechnol. Prog., 33:1160–1168, 2017     

Investigation of the source‐detector separation in near infrared spectroscopy for healthy and clinical applications

Understanding near infrared light propagation in tissue is vital for designing next generation optical brain imaging devices. Monte Carlo (MC) simulations provide a controlled mechanism to characterize and evaluate contributions of diverse near infrared spectroscopy (NIRS) sensor configurations and parameters. In this study, we developed a multilayer adult digital head model under both healthy and clinical settings and assessed light‐tissue interaction through MC simulations in terms of partial differential pathlength, mean total optical pathlength, diffuse reflectance, detector light intensity and spatial sensitivity profile of optical measurements. The model incorporated four layers: scalp, skull, cerebrospinal‐fluid and cerebral cortex with and without a customizable lesion for modeling hematoma of different sizes and depths. The effect of source‐detector separation (SDS) on optical measurements' sensitivity to brain tissue was investigated. Results from 1330 separate simulations [(4 lesion volumes × 4 lesion depths for clinical +3 healthy settings) × 7 SDS × 10 simulation = 1330)] each with 100 million photons indicated that selection of SDS is critical to acquire optimal measurements from the brain and recommended SDS to be 25 to 35 mm depending on the wavelengths to obtain optical monitoring of the adult brain function. The findings here can guide the design of future NIRS probes for functional neuroimaging and clinical diagnostic systems.      

In situ near infrared spectroscopy for analyte‐specific monitoring of glucose and ammonium in streptomyces coelicolor fermentations

There are many challenges associated with in situ collection of near infrared (NIR) spectra in a fermentation broth, particularly for highly aerated and agitated fermentations with filamentous organisms. In this study, antibiotic fermentation by the filamentous bacterium Streptomyces coelicolor was used as a model process. Partial least squares (PLS) regression models were calibrated for glucose and ammonium based on NIR spectra collected in situ. To ensure that the models were calibrated based on analyte‐specific information, semisynthetic samples were used for model calibration in addition to data from standard batches. Thereby, part of the inherent correlation between the analytes could be eliminated. The set of semisynthetic samples were generated from fermentation broth from five separate fermentations to which different amounts of glucose, ammonium, and biomass were added. This method has previously been used off line but never before in situ. The use of semisynthetic samples along with validation on an independent batch provided a critical and realistic evaluation of analyte‐specific models based on in situ NIR spectroscopy. The prediction of glucose was highly satisfactory resulting in a RMSEP of 1.1 g/L. The prediction of ammonium based on NIR spectra collected in situ was not satisfactory. A comparison with models calibrated based on NIR spectra collected off line suggested that this is caused by signal attenuation in the optical fibers in the region above 2,000 nm; a region which contains important absorption bands for ammonium. For improved predictions of ammonium in situ, it is suggested to focus efforts on enhancing the signal in that particular region. © 2009 American Institute of Chemical Engineers Biotechnol. Prog., 2010     

Near infrared spectroscopy coupled with radial basis function neural network for at-line monitoring of Lactococcus lactis subsp. fermentation

In our previous work, partial least squares (PLSs) were employed to develop the near infrared spectroscopy (NIRs) models for at-line (fast off-line) monitoring key parameters of Lactococcus lactis subsp. fermentation. In this study, radial basis function neural network (RBFNN) as a non-linear modeling method was investigated to develop NIRs models instead of PLS. A method named moving window radial basis function neural network (MWRBFNN) was applied to select the characteristic wavelength variables by using the degree approximation (Da) as criterion. Next, the RBFNN models with selected wavelength variables were optimized by selecting a suitable constant spread. Finally, the effective spectra pretreatment methods were selected by comparing the robustness of the optimum RBFNN models developed with pretreated spectra. The results demonstrated that the robustness of the optimal RBFNN models were better than the PLS models for at-line monitoring of glucose and pH of L. lactis subsp. fermentation.     

Fingerprint detection and process prediction by multivariate analysis of fed‐batch monoclonal antibody cell culture data

This work presents a sequential data analysis path, which was successfully applied to identify important patterns (fingerprints) in mammalian cell culture process data regarding process variables, time evolution and process response. The data set incorporates 116 fed‐batch cultivation experiments for the production of a Fc‐Fusion protein. Having precharacterized the evolutions of the investigated variables and manipulated parameters with univariate analysis, principal component analysis (PCA) and partial least squares regression (PLSR) are used for further investigation. The first major objective is to capture and understand the interaction structure and dynamic behavior of the process variables and the titer (process response) using different models. The second major objective is to evaluate those models regarding their capability to characterize and predict the titer production. Moreover, the effects of data unfolding, imputation of missing data, phase separation, and variable transformation on the performance of the models are evaluated. © 2015 American Institute of Chemical Engineers Biotechnol. Prog., 31:1633–1644, 2015