Gaussian graphical models explore dependence relationships between random variables, through the estimation of the corresponding inverse covariance matrices. In this paper we develop an estimator for such models appropriate for data from several graphical models that share the same variables and some of the dependence structure. In this setting, estimating a single graphical model would mask the underlying heterogeneity, while estimating separate models for each category does not take advantage of the common structure. We propose a method that jointly estimates the graphical models corresponding to the different categories present in the data, aiming to preserve the common structure, while allowing for differences between the categories. This is achieved through a hierarchical penalty that targets the removal of common zeros in the inverse covariance matrices across categories. We establish the asymptotic consistency and sparsity of the proposed estimator in the high-dimensional case, and illustrate its performance on a number of simulated networks. An application to learning semantic connections between terms from webpages collected from computer science departments is included.

A high-performance method was developed for protein secondary structure prediction based on the dual-layer support vector machine (SVM) and position-specific scoring matrices (PSSMs). SVM is a new machine learning technology that has been successfully applied in solving problems in the field of bioinformatics. The SVM's performance is usually better than that of traditional machine learning approaches. The performance was further improved by combining PSSM profiles with the SVM analysis. The PSSMs were generated from PSI-BLAST profiles, which contain important evolution information. The final prediction results were generated from the second SVM layer output. On the CB513 data set, the three-state overall per-residue accuracy, Q3, reached 75.2%, while segment overlap (SOV) accuracy increased to 80.0%. On the CB396 data set, the Q3 of our method reached 74.0% and the SOV reached 78.1%. A web server utilizing the method has been constructed and is available at http://www.bioinfo.tsinghua.edu.cn/pmsvm.

Abstract:In the present contribution, a complexity measure is proposed to assess surface 1 electromyography (EMG) in the study of muscle fatigue during sustained, isometric muscle 2 contractions. Approximate entropy (ApEn) is believed to provide quantitative information about the 3 complexity of experimental data that is often corrupted with noise, short data-length, and in many cases, 4 has inherent dynamics that exhibit both deterministic and stochastic behaviors. We developed an 5 improved ApEn measure, i.e., fuzziness approximate entropy (fApEn), which utilizes the fuzzy 6 membership function to define the vectors' similarity. Tests were conducted on independent, identically 7 distributed (i.i.d.) Gaussian and uniform noises, a chirp signal, MIX processes, and Rossler, and Henon 8 maps. Compared with the standard ApEn, the fApEn showed better monotonicity, relative consistency, 9 and more robustness to noise when characterizing signals with different complexities. Performance 10 analysis on experimental EMG signals demonstrated that the fApEn significantly decreased during the 11 development of muscle fatigue, which is a similar trend to that of the mean frequency (MNF) of the 12 EMG signal, while the standard ApEn failed to detect this change. Moreover, the fApEn is more 13 sensitive to muscle fatigue than MNF with a larger linear regression slope (significant value p=0.0213). 14 The results suggest that the fApEn of an EMG signal may potentially become a new reliable method for 15 muscle fatigue assessment and be applicable to other short noisy physiological signal analysis. 16 17

Previous works have resulted in some practical achievements for mechanomyogram (MMG) to control powered prostheses. This work presents the investigation of classifying the hand motion using MMG signals for multifunctional prosthetic control. MMG is thought to reflect the intrinsic mechanical activity of muscle from the lateral oscillations of fibers during contraction. However, external mechanical noise sources such as a movement artifact are known to cause considerable interference to MMG, compromising the classification accuracy. To solve this noise problem, we proposed a new scheme to extract robust MMG features by the integration of the wavelet packet transform (WPT), singular value decomposition (SVD) and a feature selection technique based on distance evaluation criteria for the classification of hand motions. The WPT was first adopted to provide an effective time-frequency representation of non-stationary MMG signals. Then, the SVD and the distance evaluation technique were utilized to extract and select the optimal feature representing the hand motion patterns from the MMG time-frequency representation matrix. Experimental results of 12 subjects showed that four different motions of the forearm and hand could be reliably differentiated using the proposed method when two channels of MMG signals were used. Compared with three previously reported time-frequency decomposition methods, i.e. short-time Fourier transform, stationary wavelet transform and S-transform, the proposed classification system gave the highest average classification accuracy up to 89.7%. The results indicated that MMG could potentially serve as an alternative source of electromyogram for multifunctional prosthetic control using the proposed classification method.