Learning the kernel for rare variant genetic association test.

Introduction: Compared to Genome-Wide Association Studies (GWAS) for common variants, single-marker association analysis for rare variants is underpowered. Set-based association analyses for rare variants are powerful tools that capture some of the missing heritability in trait association studies. Methods: We extend the convex-optimized SKAT (cSKAT) test set procedure which learns from data the optimal convex combination of kernels, to the full Generalised Linear Model (GLM) setting with arbitrary non-genetic covariates. We call this extended cSKAT (ecSKAT) and show that the resulting optimization problem is a quadratic programming problem that can be solved with no additional cost compared to cSKAT. Results: We show that a modified objective is related to an upper bound for the p-value through a decreasing exponential term in the objective function, indicating that optimizing this objective function is a principled way of learning the combination of kernels. We evaluate the performance of the proposed method on continuous and binary traits using simulation studies and illustrate its application using UK Biobank Whole Exome Sequencing data on hand grip strength and systemic lupus erythematosus rare variant association analysis. Discussion: Our proposed ecSKAT method enables correcting for important confounders in association studies such as age, sex or population structure for both quantitative and binary traits. Simulation studies showed that ecSKAT can recover sensible weights and achieve higher power across different sample sizes and misspecification settings. Compared to the burden test and SKAT method, ecSKAT gives a lower p-value for the genes tested in both quantitative and binary traits in the UKBiobank cohort.

Sub-band target alignment common spatial pattern in brain-computer interface.

BACKGROUND AND OBJECTIVE: In the brain computer interface (BCI) field, using sub-band common spatial pattern (SBCSP) and filter bank common spatial pattern (FBCSP) can improve the accuracy of classification by selection a specific frequency band. However, in the cross-subject classification, due to the individual differences between different subjects, the performance is limited. METHODS: This paper introduces the idea of transfer learning and presents the sub-band target alignment common spatial pattern (SBTACSP) method and applies it to the cross-subject classification of motor imagery (MI) EEG signals. First, the EEG signals are bandpass-filtered into multiple frequency bands (sub-band filtering). Subsequently, the source domain trails are aligned into the target domain space in each frequency band. The CSP algorithm is then employed to extract features among which more representative features are selected by the minimum redundancy maximum relevance (mRMR) approach from each sub-band. Then the features of all sub-bands are fused. Finally, conventional linear discriminant analysis (LDA) algorithm is used for MI classification. RESULTS: Our method is evaluated on Datasets Ⅱa and Ⅱb of the BCI Competition Ⅳ. Compared with six state-of-the-art algorithms, the proposed SBTACSP method performed relatively the best and achieved a mean classification accuracy of 75.15% and 66.85% in cross-subject classification of Datasets Ⅱa and Ⅱb respectively. CONCLUSION: Therefore, the combination of sub-band filtering and transfer learning achieves superior classification performance compared to either one. The proposed algorithms will greatly promote the practical application of MI based BCIs.

A novel miniature robotic guidance device for stereotactic neurosurgical interventions: preliminary experience with the iSYS1 robot.

OBJECTIVE Robotic devices have recently been introduced in stereotactic neurosurgery in order to overcome the limitations of frame-based and frameless techniques in terms of accuracy and safety. The aim of this study is to evaluate the feasibility and accuracy of the novel, miniature, iSYS1 robotic guidance device in stereotactic neurosurgery. METHODS A preclinical phantom trial was conducted to compare the accuracy and duration of needle positioning between the robotic and manual technique in 162 cadaver biopsies. Second, 25 consecutive cases of tumor biopsies and intracranial catheter placements were performed with robotic guidance to evaluate the feasibility, accuracy, and duration of system setup and application in a clinical setting. RESULTS The preclinical phantom trial revealed a mean target error of 0.6 mm (range 0.1-0.9 mm) for robotic guidance versus 1.2 mm (range 0.1-2.6 mm) for manual positioning of the biopsy needle (p < 0.001). The mean duration was 2.6 minutes (range 1.3-5.5 minutes) with robotic guidance versus 3.7 minutes (range 2.0-10.5 minutes) with manual positioning (p < 0.001). Clinical application of the iSYS1 robotic guidance device was feasible in all but 1 case. The median real target error was 1.3 mm (range 0.2-2.6 mm) at entry and 0.9 mm (range 0.0-3.1 mm) at the target point. The median setup and instrument positioning times were 11.8 minutes (range 4.2-26.7 minutes) and 4.9 minutes (range 3.1-14.0 minutes), respectively. CONCLUSIONS According to the preclinical data, application of the iSYS1 robot can significantly improve accuracy and reduce instrument positioning time. During clinical application, the robot proved its high accuracy, short setup time, and short instrument positioning time, as well as demonstrating a short learning curve.

Fast Gaussian kernel learning for classification tasks based on specially structured global optimization.

For a practical pattern classification task solved by kernel methods, the computing time is mainly spent on kernel learning (or training). However, the current kernel learning approaches are based on local optimization techniques, and hard to have good time performances, especially for large datasets. Thus the existing algorithms cannot be easily extended to large-scale tasks. In this paper, we present a fast Gaussian kernel learning method by solving a specially structured global optimization (SSGO) problem. We optimize the Gaussian kernel function by using the formulated kernel target alignment criterion, which is a difference of increasing (d.i.) functions. Through using a power-transformation based convexification method, the objective criterion can be represented as a difference of convex (d.c.) functions with a fixed power-transformation parameter. And the objective programming problem can then be converted to a SSGO problem: globally minimizing a concave function over a convex set. The SSGO problem is classical and has good solvability. Thus, to find the global optimal solution efficiently, we can adopt the improved Hoffman's outer approximation method, which need not repeat the searching procedure with different starting points to locate the best local minimum. Also, the proposed method can be proven to converge to the global solution for any classification task. We evaluate the proposed method on twenty benchmark datasets, and compare it with four other Gaussian kernel learning methods. Experimental results show that the proposed method stably achieves both good time-efficiency performance and good classification performance.