Evaluating the performance of machine learning methods and variable selection methods for predicting difficult-to-measure traits in Holstein dairy cattle using milk infrared spectral data

Fourier-transform infrared (FTIR) spectroscopy is a powerful high-throughput phenotyping tool for predicting traits that are expensive and difficult to measure in dairy cattle. Calibration equations are often developed using standard methods, such as partial least squares (PLS) regression. Methods that employ penalization, rank-reduction, and variable selection, as well as being able to model the nonlinear relations between phenotype and FTIR, might offer improvements in predictive ability and model robustness. This study aimed to compare the predictive ability of 2 machine learning methods, namely random forest (RF) and gradient boosting machine (GBM), and penalized regression against PLS regression for predicting 3 phenotypes differing in terms of biological meaning and relationships with milk composition (i.e., phenotypes measurable directly and not directly in milk, reflecting different biological processes which can be captured using milk spectra) in Holstein-Friesian cattle under 2 cross-validation scenarios. The data set comprised phenotypic information from 471 Holstein-Friesian cows, and 3 target phenotypes were evaluated: (1) body condition score (BCS), (2) blood β-hydroxybutyrate (BHB, mmol/L), and (3) κ-casein expressed as a percentage of nitrogen (κ-CN, % N). The data set was split considering 2 cross-validation scenarios: samples-out random in which the population was randomly split into 10-folds (8-folds for training and 1-fold for validation and testing); and herd/date-out in which the population was randomly assigned to training (70% herd), validation (10%), and testing (20% herd) based on the herd and date in which the samples were collected. The random grid search was performed using the training subset for the hyperparameter optimization and the validation set was used for the generalization of prediction error. The trained model was then used to assess the final prediction in the testing subset. The grid search for penalized regression evidenced that the elastic net (EN) was the best regularization with increase in predictive ability of 5%. The performance of PLS (standard model) was compared against 2 machine learning techniques and penalized regression using 2 cross-validation scenarios. Machine learning methods showed a greater predictive ability for BCS (0.63 for GBM and 0.61 for RF), BHB (0.80 for GBM and 0.79 for RF), and κ-CN (0.81 for GBM and 0.80 for RF) in samples-out cross-validation. Considering a herd/date-out cross-validation these values were 0.58 (GBM and RF) for BCS, 0.73 (GBM and RF) for BHB, and 0.77 (GBM and RF) for κ-CN. The GBM model tended to outperform other methods in predictive ability around 4%, 1%, and 7% for EN, RF, and PLS, respectively. The prediction accuracies of the GBM and RF models were similar, and differed statistically from the PLS model in samples-out random cross-validation. Although, machine learning techniques outperformed PLS in herd/date-out cross-validation, no significant differences were observed in terms of predictive ability due to the large standard deviation observed for predictions. Overall, GBM achieved the highest accuracy of FTIR-based prediction of the different phenotypic traits across the cross-validation scenarios. These results indicate that GBM is a promising method for obtaining more accurate FTIR-based predictions for different phenotypes in dairy cattle.     

Predicting bull fertility using genomic data and biological information

The genomic prediction of unobserved genetic values or future phenotypes for complex traits has revolutionized agriculture and human medicine. Fertility traits are undoubtedly complex traits of great economic importance to the dairy industry. Although genomic prediction for improved cow fertility has received much attention, bull fertility largely has been ignored. The first aim of this study was to investigate the feasibility of genomic prediction of sire conception rate (SCR) in US Holstein dairy cattle. Standard genomic prediction often ignores any available information about functional features of the genome, although it is believed that such information can yield more accurate and more persistent predictions. Hence, the second objective was to incorporate prior biological information into predictive models and evaluate their performance. The analyses included the use of kernel-based models fitting either all single nucleotide polymorphisms (SNP; 55K) or only markers with presumed functional roles, such as SNP linked to Gene Ontology or Medical Subject Heading terms related to male fertility, or SNP significantly associated with SCR. Both single- and multikernel models were evaluated using linear and Gaussian kernels. Predictive ability was evaluated in 5-fold cross-validation. The entire set of SNP exhibited predictive correlations around 0.35. Neither Gene Ontology nor Medical Subject Heading gene sets achieved predictive abilities higher than their counterparts using random sets of SNP. Notably, kernel models fitting significant SNP achieved the best performance with increases in accuracy up to 5% compared with the standard whole-genome approach. Models fitting Gaussian kernels outperformed their counterparts fitting linear kernels irrespective of the set of SNP. Overall, our findings suggest that genomic prediction of bull fertility is feasible in dairy cattle. This provides potential for accurate genome-guided decisions, such as early culling of bull calves with low SCR predictions. In addition, exploiting nonlinear effects through the use of Gaussian kernels together with the incorporation of relevant markers seems to be a promising alternative to the standard approach. The inclusion of gene set results into prediction models deserves further research.     

The role of motor evoked potentials during surgery for intramedullary spinal cord tumors

OBJECTIVE: This is a prospective study of the methodology and clinical applications of motor evoked potentials (MEPs) during surgery for intramedullary spinal cord tumors. METHODS: Transcranial electrical stimulation was used to activate corticospinal motoneurons, and the traveling waves of the spinal cord were recorded through catheter-electrodes placed epi- or subdurally. Intraoperative MEP monitoring was performed in 32 consecutive patients (age range, 1-50 yr) undergoing resection of intramedullary spinal cord tumors. In 19 patients, MEPs were present before myelotomy (monitorable group), and in 10 patients, MEPs were absent before myelotomy (unmonitorable group). Placement of an epidural electrode was not possible in two patients, and technical problems prevented recording in one. RESULTS: MEP amplitudes decreased intraoperatively by more than 50% of baseline in three patients, all of whom had postoperative paraplegia. Two of these patients recovered within 1 week after surgery, and one remained paraplegic. None of the patients with preserved MEP amplitude (> 50%) sustained immediate significant postoperative deterioration. Motor function was significantly deteriorated 1 week after surgery in one patient in the monitorable group and in five patients in the unmonitorable group. MEP monitorability was significantly associated with good surgical outcome for adult patients (P < 0.05), although not for pediatric patients (P > 0.6). Preoperative motor status and surgical outcome were not significantly associated for the adult (P = 0.13) or pediatric groups (P > 0.4). CONCLUSION: MEP monitorability was a better predictor of functional outcome than the patient's preoperative motor status for the adult group. Significant predictors of MEP monitorability in the adult group were preoperative motor function (P < 0.01), history of no previous treatment (surgery or irradiation) (P < 0.01), and small tumor size (P < 0.05). Weak associations with monitorable MEPs existed for low-grade tumors (P = 0.09), the presence of baseline somatosensory evoked potentials (P = 0.10), and tumor pathological abnormalities (ependymoma) (P = 0.13). No associations were determined for sex (P > 0.4), associated syrinx (P > 0.3), or tumor location (P > 0.5). In the pediatric group, none of the examined factors were associated with MEP monitorability (P > 0.3). A decline of more than 50% in MEP amplitude during tumor removal should serve as a serious warning sign to the surgeon.     

A surgical case of acute three-channeled aortic dissection in Marfan syndrome

We surgically treated a 35-year-old male with acute 3-channeled aortic dissection in Marfan syndrome. He had acute type A aortic dissection, and underwent Bentall's type operation, simultaneous graft replacement of the ascending aorta and total aortic arch. Pain recurred 5 years and 9 months after the first operation. CT scan showed two adjacent false lumens in the descending aorta. The morphology of the first and second dissections was Stanford type A + B. The second dissection was acute. In the second false lumen, a re-entry formation was observed in the abdominal aorta. Because severe pain was persistent, we immediately replaced the descending aorta using a femoro-femoral partial cardiopulmonary bypass. The patient was doing well and was discharged. When pain recurs in a Marfan patient with an aortic dissection, a 3-channeled aortic dissection should be suspected, and we recommend emergency surgery.     

Recurrent type III membranoproliferative glomerulonephritis after kidney transplantation

The first well-documented case of recurrent type III membranoproliferative glomerulonephritis after kidney transplantation is reported in this article. A 48-year-old man was admitted to the hospital because of nephrotic syndrome and moderate renal failure. The renal biopsy showed double-contour images at light microscopy. Electron microscopy revealed electron-dense deposits in the mesangium and in both the subepithelial and subendothelial sides of the basement membrane. Subepithelial deposits were sometimes hump-like and produced an irregular disruption of the lamina densa. A diagnosis of type III membranoproliferative glomerulonephritis was suggested. The patient had a rapid decrease in renal function and received dialysis in 3 months. Three years later, he received a cadaveric kidney transplant, and subsequently recovered normal renal function. Proteinuria appeared after 13 months, and a biopsy of the graft demonstrated recurrence of the original disease. Seven years after transplantation, he returned to hemodialysis.     

Magnetic probe array with high sensitivity for fluctuating field

A magnetic probe array is constructed to measure precisely the spatial structure of a small fluctuating field included in a strong confinement field that varies with time. To exclude the effect of the confinement field, the magnetic probes consisting of figure-eight-wound coils are prepared. The spatial structure of the fluctuating field is obtained from a Fourier analysis of the probe signal. It is found that the probe array is more sensitive to the fluctuating field with a high mode number than that with a low mode number. An experimental demonstration of the present method is attempted using a field-reversed configuration plasma, where the fluctuating field with 0.1% of the confinement field is successfully detected.     

Chemical and electronic interface structure of spray pyrolysis deposited undoped and Al-doped ZnO thin films on a commercial Cz-Si solar cell substrate

We have studied differences in the interface between undoped and Al-doped ZnO thin films deposited on commercial Si solar cell substrates. The undoped ZnO film is significantly thicker than the Al-doped film for the same deposition time. An extended silicate-like interface is present in both samples. Transmission electron microscopy (TEM) and photoelectron spectroscopy (PES) probe the presence of a zinc silicate and several Si oxides in both cases. Although Al doping improves the conductivity of ZnO, we present evidence for Al segregation at the interface during deposition on the Si substrate and suggest the presence of considerable fixed charge near the oxidized Si interface layer. The induced distortion in the valence band, compared to that of undoped ZnO, could be responsible for considerable reduction in the solar cell performance.     

Developing an Updated Strategy for Estimating the Free-Energy Parameters in RNA Duplexes

For the last 20 years, it has been common lore that the free energy of RNA duplexes formed from canonical Watson–Crick base pairs (bps) can be largely approximated with dinucleotide bp parameters and a few simple corrective constants that are duplex independent. Additionally, the standard benchmark set of duplexes used to generate the parameters were GC-rich in the shorter duplexes and AU-rich in the longer duplexes, and the length of the majority of the duplexes ranged between 6 and 8 bps. We were curious if other models would generate similar results and whether adding longer duplexes of 17 bps would affect the conclusions. We developed a gradient-descent fitting program for obtaining free-energy parameters—the changes in Gibbs free energy (ΔG), enthalpy (ΔH), and entropy (ΔS), and the melting temperature (Tm)—directly from the experimental melting curves. Using gradient descent and a genetic algorithm, the duplex melting results were combined with the standard benchmark data to obtain bp parameters. Both the standard (Turner) model and a new model that includes length-dependent terms were tested. Both models could fit the standard benchmark data; however, the new model could handle longer sequences better. We developed an updated strategy for fitting the duplex melting data.     

A molecular dynamics study on liquid-vapor interface adsorbed by impurities

In this paper, we focus on the microscopic structure of the liquid-vapor interface and the effect of impurities such as surfactants or electrolytes in the aqueous solution on interfacial properties. We perform molecular dynamics (MD) simulations of a simple planar interface system and a bubble system in which a nanometer-size void region is maintained. In these MD systems, some static and dynamic properties of molecules near the interface are investigated and we elucidate some microscopic features of the interface. Especially, we show the difference of the dynamics of water molecules between the surfactant and electrolyte solutions. We also consider the definition of the interface at the molecular level which can treat the local and instantaneous structure, and discuss the physical properties which we calculate including the effect of molecular fluctuations of the interface.     

The Characteristic Structure of Anti‐HIV Actinohivin in Complex with Three HMTG D1 Chains of HIV‐gp120

The anti‐HIV lectin actinohivin (AH) specifically interacts with HMTG (high‐mannose‐type glycan), which is attached to the glycoprotein gp120 of HIV‐1 in a process in which the three branched mannotriose chains (D1, D2, and D3) of HMTG exhibit different binding affinities, it being estimated that that of D1 is the strongest, that of D3 is weaker, and that of D2 is undetectable. These properties have been ascribed to the stereochemical differences in linkages between the second and the third mannose residues of the three chains. In order to clarify the interaction geometry between AH and the major target D1, an X‐ray determination of the crystal structure of AH in complex with D1—which is α(1,2)mannotriose composed of three mannose (Man) residues linked together only by α(1,2) bonding—has been performed. In each of the three D1‐binding pockets of AH, two Man residues of D1 are accommodated at zones 1 and 2 in the pocket, in the same way as those found in the α(1,2)mannobiose‐bound AH crystals. However, an OMIT map shows poor densities at both ends of the two residues. This suggests the existence of positional disorder of D1 in the pocket: the two zones are each occupied by two Man residues in two different modes, with mode A involving the Man1 and Man2 residues and mode B the Man2 and Man3 residues. In each mode, D1 is stabilized by adopting a double‐bracket‐shaped conformation through C?H ??? O interactions. In mode B, however, the Man1 residue, which is the most sensitive residue to AH binding, protrudes wholly into the solvent region without contacts with AH. In mode A, in contrast, the Man3 residue interacts with the essential hydrophobic amino acid residues (Tyr and Leu conserved between the three pockets) of AH. Therefore, mode A is likely to be the one that occurs when whole HMTG is bound. In this mode, the two hydroxy groups (O3 and O4) of the Man2 residue are anchored in zone 2 by four hydrogen bonds with Asp, Asn, and Tyr residues of AH. In addition, it has been found that an isolated water molecule buried in the hydrophobic long loop bridges between Asp of AH and the hydroxy group of Man2 through hydrogen bonds. The most interesting feature is found in the interaction of the Man1 and Man3 residues with AH. All eight hydroxy groups of the two residues are completely exposed in the solvent region, whereas their hydrophobic parts make contacts with a Leu residue and two Tyr residues so that the shape of D1 and the surface of AH fit well over a wide area. These structural characteristics are potentially useful for development of AH to produce more effective antiretroviral drugs to suppress the infectious expansion of HIV/AIDS and to help expedite an end to the HIV/AIDS pandemic in the near future.