Features extraction and multi-classification of sEMG using a GPU-Accelerated GA/MLP hybrid algorithm.

BACKGROUND: Surface electromyography (sEMG) signal is the combined effect of superficial muscle EMG and neural electrical activity. In recent years, researchers did large amount of human-machine system studies by using the physiological signals as control signals. OBJECTIVE: To develop and test a new multi-classification method to improve performance of analyzing sEMG signals based on public sEMG dataset. METHODS: First, ten features were selected as candidate features. Second, a genetic algorithm (GA) was applied to select representative features from the initial ten candidates. Third, a multi-layer perceptron (MLP) classifier was trained by the selected optimal features. Last, the trained classifier was used to predict the classes of sEMG signals. A special graphics processing unit (GPU) was used to speed up the learning process. RESULTS: Experimental results show that the classification accuracy of the new method reached higher than 90%. Comparing to other previously reported results, using the new method yielded higher performance. CONCLUSIONS: The proposed features selection method is effective and the classification result is accurate. In addition, our method could have practical application value in medical prosthetics and the potential to improve robustness of myoelectric pattern recognition.

A two-dimensional matrix image based feature extraction method for classification of sEMG: A comparative analysis based on SVM, KNN and RBF-NN.

BACKGROUND: The computer mouse is an important human-computer interaction device. But patients with physical finger disability are unable to operate this device. Surface EMG (sEMG) can be monitored by electrodes on the skin surface and is a reflection of the neuromuscular activities. Therefore, we can control limbs auxiliary equipment by utilizing sEMG classification in order to help the physically disabled patients to operate the mouse. OBJECTIVE: To develop a new a method to extract sEMG generated by finger motion and apply novel features to classify sEMG. METHODS: A window-based data acquisition method was presented to extract signal samples from sEMG electordes. Afterwards, a two-dimensional matrix image based feature extraction method, which differs from the classical methods based on time domain or frequency domain, was employed to transform signal samples to feature maps used for classification. In the experiments, sEMG data samples produced by the index and middle fingers at the click of a mouse button were separately acquired. Then, characteristics of the samples were analyzed to generate a feature map for each sample. Finally, the machine learning classification algorithms (SVM, KNN, RBF-NN) were employed to classify these feature maps on a GPU. RESULTS: The study demonstrated that all classifiers can identify and classify sEMG samples effectively. In particular, the accuracy of the SVM classifier reached up to 100%. CONCLUSIONS: The signal separation method is a convenient, efficient and quick method, which can effectively extract the sEMG samples produced by fingers. In addition, unlike the classical methods, the new method enables to extract features by enlarging sample signals' energy appropriately. The classical machine learning classifiers all performed well by using these features.

Genome-wide scan reveals LEMD3 and WIF1 on SSC5 as the candidates for porcine ear size.

The quantitative trait loci (QTL) for porcine ear size was previously reported to mainly focus on SSC5 and SSC7. Recently, a missense mutation, G32E, in PPARD in the QTL interval on SSC7 was identified as the causative mutation for ear size. However, on account of the large interval of QTL, the responsible gene on SSC5 has not been identified. In this study, an intercross population was constructed from the large-eared Minzhu, an indigenous Chinese pig breed, and the Western commercial Large White pig to examine the genetic basis of ear size diversity. A GWAS was performed to detect SNPs significantly associated with ear size. Thirty-five significant SNPs defined a 10.78-Mb (30.14-40.92 Mb) region on SSC5. Further, combining linkage disequilibrium and haplotype sharing analysis, a reduced region of 3.07-Mb was obtained. Finally, by using a selective sweep analysis, a critical region of about 450-kb interval containing two annotated genes LEMD3 and WIF1 was refined in this work. Functional analysis indicated that both represent biological candidates for porcine ear size, with potential application in breeding programs. The two genes could also be used as novel references for further study of the mechanism underlying human microtia.

Genome-wide copy number variations inferred from SNP genotyping arrays using a Large White and Minzhu intercross population.

Copy number variations (CNVs) are one of the main contributors to genetic diversity in animals and are broadly distributed in the genomes of swine. Investigating the performance and evolutionary impacts of pig CNVs requires comprehensive knowledge of their structure and function within and between breeds. In the current study, 4 different programs (i.e., GADA, PennCNV, QuantiSNP, and cnvPartition) were used to analyze Porcine SNP60 genotyping data of 585 pigs from one Large White × Minzhu intercross population to detect copy number variant regions (CNVRs). Overlapping CNVRs recalled by at least 2 programs were used to construct a powerful and comprehensive CNVR map, which contained 249 CNVRs (i.e., 70 gains, 43 losses, and 136 gains/losses) and covered 26.22% of the regions in the swine genome. Ten CNVRs, representing different predicted statuses, were selected for validation via quantitative real-time PCR (QPCR); 9/10 CNVRs (i.e., 90%) were validated. When being traced back to the F0 generation, 58 events were identified in only Minzhu F0 parents and 2 events were identified in only Large White F0 parents. A series of CNVR function analyses were performed. Some of the CNVRs functions were predicted, and several interesting CNVRs for meat quality traits and hematological parameters were obtained. A comprehensive and lower false rate genome-wide CNV map was constructed for Large White and Minzhu pig genomes in this study. Our results may provide an important basis for determining the relationship between CNVRs and important qualitative and quantitative traits. In addition, it can help to further understand genetic processes in pigs.

Genome-wide association study of porcine hematological parameters in a Large White × Minzhu F2 resource population.

Hematological traits, which are important indicators of immune function in animals, have been commonly examined as biomarkers of disease and disease severity in humans and animals. Genome-wide significant quantitative trait loci (QTLs) provide important information for use in breeding programs of animals such as pigs. QTLs for hematological parameters (hematological traits) have been detected in pig chromosomes, although these are often mapped by linkage analysis to large intervals making identification of the underlying mutation problematic. Single nucleotide polymorphisms (SNPs) are the common form of genetic variation among individuals and are thought to account for the majority of inherited traits. In this study, a genome-wide association study (GWAS) was performed to detect regions of association with hematological traits in a three-generation resource population produced by intercrossing Large White boars and Minzhu sows during the period from 2007 to 2011. Illumina PorcineSNP60 BeadChip technology was used to genotype each animal and seven hematological parameters were measured (hematocrit (HCT), hemoglobin (HGB), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), mean corpuscular volume (MCV), red blood cell count (RBC) and red blood cell volume distribution width (RDW)). Data were analyzed in a three step Genome-wide Rapid Association using the Mixed Model and Regression-Genomic Control (GRAMMAR-GC) method. A total of 62 genome-wide significant and three chromosome-wide significant SNPs associated with hematological parameters were detected in this GWAS. Seven and five SNPs were associated with HCT and HGB, respectively. These SNPs were all located within the region of 34.6-36.5 Mb on SSC7. Four SNPs within the region of 43.7-47.0 Mb and fifty-five SNPs within the region of 42.2-73.8 Mb on SSC8 showed significant association with MCH and MCV, respectively. At chromosome-wide significant level, one SNP at 29.2 Mb on SSC1 and two SNPs within the region of 26.0-26.2 Mb were found to be significantly associated with RBC and RDW, respectively. Many of the SNPs were located within previously reported QTL regions and appeared to narrow down the regions compared with previously described QTL intervals. In current research, a total of seven significant SNPs were found within six candidate genes SCUBE3, KDR, TDO, IGFBP7, ADAMTS3 and AFP. In addition, the KIT gene, which has been previously reported to relate to hematological parameters, was located within the region significantly associated with MCH and MCV and could be a candidate gene. These results of this study may lead to a better understanding of the molecular mechanisms of hematological parameters in pigs.

Data mining in networks of differentially expressed genes during sow pregnancy.

Small to moderate gains in Pig fertility can mean large returns in overall efficiency, and developing methods to improve it is highly desirable. High fertility rates depend on completion of successful pregnancies. To understand the molecular signals associated with pregnancy in sows, expression profiling experiments were conducted to identify differentially expressed genes in ovary and myometrium at different pregnancy periods using the Affymetrix Porcine GeneChip(TM). A total of 974, 1800, 335 and 710 differentially expressed transcripts were identified in the myometrium during early pregnancy (EP) and late pregnancy (LP), and in the ovary during EP and LP, respectively. Self-Organizing Map (SOM) clusters indicated the differentially expressed genes belonged to 7 different functional groups. Based on BLASTX searches and Gene Ontology (GO) classifications, 129 unique genes closely related to pregnancy showed differential expression patterns. GO analysis also indicated that there were 21 different molecular function categories, 20 different biological process categories, and 8 different cellular component categories of genes differentially expressed during sow pregnancy. Gene regulatory network reconstruction provided us with an interaction model of known genes such as insulin-like growth factor 2 (IGF2) gene, estrogen receptor (ESR) gene, retinol-binding protein-4 (RBP4) gene, and several unknown candidate genes related to reproduction. Several pitch point genes were selected for association study with reproduction traits. For instance, DPPA5 g.363 T>C was found to associate with litter born weight at later parities in Beijing Black pigs significantly (p < 0.05). Overall, this study contributes to elucidating the mechanism underlying pregnancy processes, which maybe provide valuable information for pig reproduction improvement.

Genome-wide association analysis of meat quality traits in a porcine Large White × Minzhu intercross population.

Pork quality is an economically important trait and one of the main selection criteria for breeding in the swine industry. In this genome-wide association study (GWAS), 455 pigs from a porcine Large White × Minzhu intercross population were genotyped using the Illumina PorcineSNP60K Beadchip, and phenotyped for intramuscular fat content (IMF), marbling, moisture, color L*, color a*, color b* and color score in the longissimus muscle (LM). Association tests between each trait and the SNPs were performed via the Genome Wide Rapid Association using the Mixed Model and Regression-Genomic Control (GRAMMAR-GC) approach. From the Ensembl porcine database, SNP annotation was implemented using Sus scrofa Build 9. A total of 45 SNPs showed significant association with one or multiple meat quality traits. Of the 45 SNPs, 36 were located on SSC12. These significantly associated SNPs aligned to or were in close approximation to previously reported quantitative trait loci (QTL) and some were located within introns of previously reported candidate genes. Two haplotype blocks ASGA0100525-ASGA0055225-ALGA0067099-MARC0004712-DIAS0000861, and ASGA0085522-H3GA0056170 were detected in the significant region. The first block contained the genes MYH1, MYH2 and MYH4. A SNP (ASGA0094812) within an intron of the USP43 gene was significantly associated with five meat quality traits. The present results effectively narrowed down the associated regions compared to previous QTL studies and revealed haplotypes and candidate genes on SSC12 for meat quality traits in pigs.

Simulation study on the efficiencies of MOET nucleus breeding schemes applying marker assisted selection in dairy cattle.

Advantages of breeding schemes using genetic marker information and/or multiple ovulation and embryo transfer (MOET) technology over the traditional approach were extensively evaluated through simulation. Milk yield was the trait of interest and QTL was the genetic marker utilized. Eight dairy cattle breeding scenarios were considered, i.e., traditional progeny testing breeding scheme (denoted as STANPT), GASPT scheme including a pre-selection of young bulls entering progeny testing based on their own QTL information, MOETPT scheme using MOET technology to generate young bulls and a selection of young bulls limited within the full-sib family, GAMOPT scheme adopting both QTL pre-selection and MOET technology, COMBPT scheme using a mixed linear model which considered QTL genotype instead of the BLUP model in GAMOPT, and three non-progeny testing schemes, i.e. the MOET, GAMO and COMB schemes, corresponding to MOETPT, GAMOPT and COMBPT with progeny testing being part of the system. Animals were selected based on their breeding value which was estimated under an animal model framework. Sequential selection over 17 years was performed in the simulations and 30 replicates were designed for each scenario. The influences of using QTL information and MOET technology on favorable QTL allele frequency, true breeding values, polygenetic breeding values and the accumulated genetic superiority were extensively evaluated, for five different populations including active sires, lactating cows, bull dams, bull sires, and young bulls. The results showed that the combined schemes significantly outperformed other approaches wherein accumulated true breeding value progressed. The difference between schemes exclusively using QTL information or MOET technology was not significant. The STANPT scheme was the least efficient among the 8 schemes. The schemes using MOET technology had a higher polygenetic response than others in the 17th year. The increases of frequency of the favorable QTL allele varied more greatly across the 3 male groups than in the lactating cows group. The accumulated genetic superiorities of the GASPT scheme, MOETPT scheme, GAMOPT scheme, COMBPT scheme, MOET scheme, GAMO scheme and COMB scheme over the STANPT scheme were 8.42%, 3.59%, 14.58%, 18.54%, 4.12%, 14.12%, 16.50% in active sires and 2.70%, 5.00%, 11.05%, 12.78%, 7.51%, 17.12%, 25.38% in lactating cows.

[Biological evolution and specialization].

There were some disputes about the concept and mechanism of biological evolution. This paper tried to give more explanations on the key concepts. The biological adaptability was distinguished into two different concepts: biological evolution and specialization. The former was defined as the process of biologically gradual evolvement, and the latter was considered as the process of species formation at horizontal development. Moreover, a new conceptual framework was applied to the popular biological theories known by people, and the previous research results or discoveries were explained over again.