Feature extraction of particle morphologies of pharmaceutical excipients from scanning electron microscope images using convolutional neural networks.

Scanning electron microscopy (SEM) images are the most widely used tool for evaluating particle morphology; however, quantitative evaluation using SEM images is time-consuming and often neglected. In this study, we aimed to extract features related to particle morphology of pharmaceutical excipients from SEM images using a convolutional neural network (CNN). SEM images of 67 excipients were acquired and used as models. A classification CNN model of the excipients was constructed based on the SEM images. Further, features were extracted from the middle layer of this CNN model, and the data was compressed to two dimensions using uniform manifold approximation and projection. Lastly, hierarchical clustering analysis (HCA) was performed to categorize the excipients into several clusters and identify similarities among the samples. The classification CNN model showed high accuracy, allowing each excipient to be identified with a high degree of accuracy. HCA revealed that the 67 excipients were classified into seven clusters. Additionally, the particle morphologies of excipients belonging to the same cluster were found to be very similar. These results suggest that CNN models are useful tools for extracting information and identifying similarities among the particle morphologies of excipients.

Classification of scanning electron microscope images of pharmaceutical excipients using deep convolutional neural networks with transfer learning.

Convolutional Neural Networks (CNNs) are image analysis techniques that have been applied to image classification in various fields. In this study, we applied a CNN to classify scanning electron microscopy (SEM) images of pharmaceutical raw material powders to determine if a CNN can evaluate particle morphology. We tested 10 pharmaceutical excipients with widely different particle morphologies. SEM images for each excipient were acquired and divided into training, validation, and test sets. Classification models were constructed by applying transfer learning to pretrained CNN models such as VGG16 and ResNet50. The results of a 5-fold cross-validation showed that the classification accuracy of the CNN model was sufficiently high using either pretrained model and that the type of excipient could be classified with high accuracy. The results suggest that the CNN model can detect differences in particle morphology, such as particle size, shape, and surface condition. By applying Grad-CAM to the constructed CNN model, we succeeded in finding particularly important regions in the particle image of the excipients. CNNs have been found to have the potential to be applied to the identification and characterization of raw material powders for pharmaceutical development.

MDGRAPE-4: a special-purpose computer system for molecular dynamics simulations.

We are developing the MDGRAPE-4, a special-purpose computer system for molecular dynamics (MD) simulations. MDGRAPE-4 is designed to achieve strong scalability for protein MD simulations through the integration of general-purpose cores, dedicated pipelines, memory banks and network interfaces (NIFs) to create a system on chip (SoC). Each SoC has 64 dedicated pipelines that are used for non-bonded force calculations and run at 0.8 GHz. Additionally, it has 65 Tensilica Xtensa LX cores with single-precision floating-point units that are used for other calculations and run at 0.6 GHz. At peak performance levels, each SoC can evaluate 51.2 G interactions per second. It also has 1.8 MB of embedded shared memory banks and six network units with a peak bandwidth of 7.2 GB s(-1) for the three-dimensional torus network. The system consists of 512 (8×8×8) SoCs in total, which are mounted on 64 node modules with eight SoCs. The optical transmitters/receivers are used for internode communication. The expected maximum power consumption is 50 kW. While MDGRAPE-4 software has still been improved, we plan to run MD simulations on MDGRAPE-4 in 2014. The MDGRAPE-4 system will enable long-time molecular dynamics simulations of small systems. It is also useful for multiscale molecular simulations where the particle simulation parts often become bottlenecks.

Accuracy of three-dimensional finite element modeling using two different dental cone beam computed tomography systems.

The aim of this study is to evaluate the accuracy of dental ceramic object three dimensional (3D) finite element model constructed directly from two different dental cone beam computed tomography (CT) systems. CT scanned one 10.0 × 10.0 × 20.0 mm block and one 8.0 × 10.0 × 40.0 mm block of an 8-step wedge. All 3D finite element (FE) models were created from CT images. Each 3D FE model measured the length of the directions X, Y, and Z that corresponded to an original specimen using the measurement function between two points on the Mechanical Finder software package. The measurements and practical value were compared with the CT image and the accuracy of the reproduced measurements was examined. No significant differences were found between Alphard-3030 on the Z axis and ProMax 3D on the Y axis of the block. In addition, there were also no significant differences observed between Alphard-3030 on the Y axis and ProMax 3D on the X axis compared with Alphard-3030 on the Z axis and ProMax 3D on the Y axis for the step-wedge. The results suggest that measurement of the dimensions of cone beam CT images could be useful in applications where both good reproducibility and accuracy of FE models are required.

Association of a polymorphism of BTN2A1 with myocardial infarction in East Asian populations.

OBJECTIVE: We have performed a genome-wide association study (GWAS) to identify genetic variants that confer susceptibility to myocardial infarction (MI) in Japanese and Korean populations. METHODS: A total of 17,447 Japanese or Korean individuals from four independent subject panels was examined. Japanese subject panels A, B, and C comprised 134 individuals with MI and 137 controls, 1431 individuals with MI and 3161 controls, and 643 individuals with MI and 1347 controls, respectively, whereas the Korean population comprised 1880 individuals with MI and 8714 controls. A GWAS for MI was performed in Japanese subject panel A with the use of the Affymetrix GeneChip Human Mapping 500K Array Set. RESULTS: Seventy single nucleotide polymorphisms (SNPs) significantly (P<1.0×10(-7)) associated with MI by the GWAS were examined further in Japanese subject panel B, revealing two SNPs (rs6929846 of BTN2A1, rs2569512 of ILF3) to be significantly (P<0.0007) associated with MI. The rs6929846 SNP of BTN2A1, but not rs2569512 of ILF3, was also significantly associated with MI in Japanese subject panel C. However, the association of neither rs6929846 nor rs2569512 with MI was replicated in the Korean population. CONCLUSION: BTN2A1 may be a susceptibility gene for MI in Japanese individuals.

Development of 3D CAD/FEM Analysis System for Natural Teeth and Jaw Bone Constructed from X-Ray CT Images.

A three-dimensional finite element model of the lower first premolar, with the three layers of enamel, dentin, and pulp, and the mandible, with the two layers of cortical and cancellous bones, was directly constructed from noninvasively acquired CT images. This model was used to develop a system to analyze the stresses on the teeth and supporting bone structure during occlusion based on the finite element method and to examine the possibility of mechanical simulation.

Identification of CELSR1 as a susceptibility gene for ischemic stroke in Japanese individuals by a genome-wide association study.

OBJECTIVE: We have performed a genome-wide association study (GWAS) to identify genetic variants that confer susceptibility to ischemic stroke. METHODS: A total of 6341 individuals from three independent populations was examined. Subject panel A comprised 131 individuals with ischemic stroke and 135 controls; subject panel B comprised 790 individuals with ischemic stroke and 3435 controls; and subject panel C comprised 71 individuals with ischemic stroke and 1779 controls. A GWAS for ischemic stroke was performed in subject panel A with the use of the GeneChip Human Mapping 500K Array Set (Affymetrix). RESULTS: The relation of 100 single nucleotide polymorphisms (SNPs) selected by the GWAS to ischemic stroke was examined in 705 subjects with ischemic stroke and 3426 controls selected from subject panel B. Three SNPs (rs1671021 of LLGL2, rs9615362 of CELSR1, and rs753307 of RUVBL2) were significantly (P<0.05) associated with ischemic stroke. After DNA sequencing of linkage disequilibrium blocks containing these SNPs, three tag SNPs (rs6007897 of CELSR1, rs1671021 of LLGL2, and rs1062708 of RUVBL2) and a nonsynonymous SNP (rs4044210 of CELSR1) were examined for their relation to ischemic stroke in subject panels B and C. Both rs6007897 (A-->G, Thr2268Ala) and rs4044210 (A-->G, Ile2107Val) of CELSR1 as well as rs1671021 (T-->C, Phe479Leu) of LLGL2 were significantly associated with ischemic stroke in subject panel B. The rs6007897 and rs4044210 polymorphisms of CELSR1 were also significantly associated with ischemic stroke in subject panel C. CONCLUSION: CELSR1 is a susceptibility gene for ischemic stroke in Japanese individuals, although the functional relevance of the identified SNPs was not determined.

Microchromosomal deletions involving SCN1A and adjacent genes in severe myoclonic epilepsy in infancy.

PURPOSE: Genetic abnormalities of the gene encoding alpha1 subunit of the sodium channel (SCN1A), which can be detected by direct sequencing, are present in more than 60% of patients with severe myoclonic epilepsy in infancy (SMEI) or its borderline phenotype (SMEB). Microchromosomal deletions have been recently reported as additional causes of SMEI. This study examines whether such microdeletions are associated with SMEI as well as with SMEB. METHODS: We recruited patients with SMEI (n = 35) and SMEB (n = 34), who were confirmed previously to have no mutations of SCN1A by direct sequencing. Microdeletions were sought by multiplex ligation-dependent probe amplification (MLPA), and then confirmed and characterized by fluorescence in situ hybridization (FISH) and array-based comparative genomic hybridization (aCGH), respectively. RESULTS: Heterozygous multiple exonic deletions were identified in 7/35 SMEI patients (20%) and 0/34 SMEB patients (0%), with a net frequency of 10.1% (7/69 patients). Deletions were confirmed by FISH and aCGH analysis. The concomitant deletions of adjacent genes were revealed by aCGH. None of the parents who agreed to undergo the analysis had such deletions suggesting that the deletions were de novo. The phenotypes of patients with the deletions were indistinguishable from those of SMEI resulting from point mutations. DISCUSSION: Our findings indicate that microchromosomal deletion, often involving not only SCN1A but also several adjacent genes, is associated with core SMEI. As microchromosomal deletion cannot be anticipated by the phenotypes or detected by conventional methods, genetic abnormalities in SMEI should be carefully sought by techniques that can detect microdeletions.

Integrative genome-wide expression analysis bears evidence of estrogen receptor-independent transcription in heregulin-stimulated MCF-7 cells.

Heregulin beta-1 (HRG) is an extracellular ligand that activates mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3-OH kinase (PI3K)/Akt signaling pathways through ErbB receptors. MAPK and Akt have been shown to phosphorylate the estrogen receptor (ER) at Ser-118 and Ser-167, respectively, thereby mimicking the effects of estrogenic activity such as estrogen responsive element (ERE)-dependent transcription. In the current study, integrative analysis was performed using two tiling array platforms, comprising histone H3 lysine 9 (H3K9) acetylation and RNA mapping, together with array comparative genomic hybridization (CGH) analysis in an effort to identify HRG-regulated genes in ER-positive MCF-7 breast cancer cells. Through application of various threshold settings, 333 (326 up-regulated and 7 down-regulated) HRG-regulated genes were detected. Prediction of upstream transcription factors (TFs) and pathway analysis indicated that 21% of HRG-induced gene regulation may be controlled by the MAPK cascade, while only 0.6% of the gene expression is controlled by ERE. A comparison with previously reported estrogen (E2)-regulated gene expression data revealed that only 12 common genes were identified between the 333 HRG-regulated (3.6%) and 239 E2-regulated (5.0%) gene groups. However, with respect to enriched upstream TFs, 4 common TFs were identified in the 14 HRG-regulated (28.6%) and 13 E2-regulated (30.8%) gene groups. These results indicated that while E2 and HRG may induce common TFs, the regulatory mechanisms that govern HRG- and E2-induced gene expression differ.