Laplacian mixture modeling for network analysis and unsupervised learning on graphs.

Laplacian mixture models identify overlapping regions of influence in unlabeled graph and network data in a scalable and computationally efficient way, yielding useful low-dimensional representations. By combining Laplacian eigenspace and finite mixture modeling methods, they provide probabilistic or fuzzy dimensionality reductions or domain decompositions for a variety of input data types, including mixture distributions, feature vectors, and graphs or networks. Provable optimal recovery using the algorithm is analytically shown for a nontrivial class of cluster graphs. Heuristic approximations for scalable high-performance implementations are described and empirically tested. Connections to PageRank and community detection in network analysis demonstrate the wide applicability of this approach. The origins of fuzzy spectral methods, beginning with generalized heat or diffusion equations in physics, are reviewed and summarized. Comparisons to other dimensionality reduction and clustering methods for challenging unsupervised machine learning problems are also discussed.

Design and characterization of a nanopore-coupled polymerase for single-molecule DNA sequencing by synthesis on an electrode array.

Scalable, high-throughput DNA sequencing is a prerequisite for precision medicine and biomedical research. Recently, we presented a nanopore-based sequencing-by-synthesis (Nanopore-SBS) approach, which used a set of nucleotides with polymer tags that allow discrimination of the nucleotides in a biological nanopore. Here, we designed and covalently coupled a DNA polymerase to an α-hemolysin (αHL) heptamer using the SpyCatcher/SpyTag conjugation approach. These porin-polymerase conjugates were inserted into lipid bilayers on a complementary metal oxide semiconductor (CMOS)-based electrode array for high-throughput electrical recording of DNA synthesis. The designed nanopore construct successfully detected the capture of tagged nucleotides complementary to a DNA base on a provided template. We measured over 200 tagged-nucleotide signals for each of the four bases and developed a classification method to uniquely distinguish them from each other and background signals. The probability of falsely identifying a background event as a true capture event was less than 1.2%. In the presence of all four tagged nucleotides, we observed sequential additions in real time during polymerase-catalyzed DNA synthesis. Single-polymerase coupling to a nanopore, in combination with the Nanopore-SBS approach, can provide the foundation for a low-cost, single-molecule, electronic DNA-sequencing platform.

Real-time single-molecule electronic DNA sequencing by synthesis using polymer-tagged nucleotides on a nanopore array.

DNA sequencing by synthesis (SBS) offers a robust platform to decipher nucleic acid sequences. Recently, we reported a single-molecule nanopore-based SBS strategy that accurately distinguishes four bases by electronically detecting and differentiating four different polymer tags attached to the 5'-phosphate of the nucleotides during their incorporation into a growing DNA strand catalyzed by DNA polymerase. Further developing this approach, we report here the use of nucleotides tagged at the terminal phosphate with oligonucleotide-based polymers to perform nanopore SBS on an α-hemolysin nanopore array platform. We designed and synthesized several polymer-tagged nucleotides using tags that produce different electrical current blockade levels and verified they are active substrates for DNA polymerase. A highly processive DNA polymerase was conjugated to the nanopore, and the conjugates were complexed with primer/template DNA and inserted into lipid bilayers over individually addressable electrodes of the nanopore chip. When an incoming complementary-tagged nucleotide forms a tight ternary complex with the primer/template and polymerase, the tag enters the pore, and the current blockade level is measured. The levels displayed by the four nucleotides tagged with four different polymers captured in the nanopore in such ternary complexes were clearly distinguishable and sequence-specific, enabling continuous sequence determination during the polymerase reaction. Thus, real-time single-molecule electronic DNA sequencing data with single-base resolution were obtained. The use of these polymer-tagged nucleotides, combined with polymerase tethering to nanopores and multiplexed nanopore sensors, should lead to new high-throughput sequencing methods.

Visualization of pelvic floor reflex and voluntary contractions.

Visualization of the geometric deformation and associated displacement patterns of tubular abdominal organs to mechanical stimuli provides a quantitative measure that is useful in modeling their elastic properties. The origin of the stimulus may be the result of direct and voluntary muscle contraction or in response to a triggered reflex activity. Using trans-perineal 2D ultrasound imaging we examined the characteristics of deformation and displacement of these organs in response to voluntary activity, contraction, straining, and fast reflex responses to stimuli such as coughing. The relative time sequence in movement was examined by serially segmenting the outline of these structures and mapping their temporal characteristics.

Managing biomedical image metadata for search and retrieval of similar images.

Radiology images are generally disconnected from the metadata describing their contents, such as imaging observations ("semantic" metadata), which are usually described in text reports that are not directly linked to the images. We developed a system, the Biomedical Image Metadata Manager (BIMM) to (1) address the problem of managing biomedical image metadata and (2) facilitate the retrieval of similar images using semantic feature metadata. Our approach allows radiologists, researchers, and students to take advantage of the vast and growing repositories of medical image data by explicitly linking images to their associated metadata in a relational database that is globally accessible through a Web application. BIMM receives input in the form of standard-based metadata files using Web service and parses and stores the metadata in a relational database allowing efficient data query and maintenance capabilities. Upon querying BIMM for images, 2D regions of interest (ROIs) stored as metadata are automatically rendered onto preview images included in search results. The system's "match observations" function retrieves images with similar ROIs based on specific semantic features describing imaging observation characteristics (IOCs). We demonstrate that the system, using IOCs alone, can accurately retrieve images with diagnoses matching the query images, and we evaluate its performance on a set of annotated liver lesion images. BIMM has several potential applications, e.g., computer-aided detection and diagnosis, content-based image retrieval, automating medical analysis protocols, and gathering population statistics like disease prevalences. The system provides a framework for decision support systems, potentially improving their diagnostic accuracy and selection of appropriate therapies.

Automated retrieval of CT images of liver lesions on the basis of image similarity: method and preliminary results.

PURPOSE: To develop a system to facilitate the retrieval of radiologic images that contain similar-appearing lesions and to perform a preliminary evaluation of this system with a database of computed tomographic (CT) images of the liver and an external standard of image similarity. MATERIALS AND METHODS: Institutional review board approval was obtained for retrospective analysis of deidentified patient images. Thereafter, 30 portal venous phase CT images of the liver exhibiting one of three types of liver lesions (13 cysts, seven hemangiomas, 10 metastases) were selected. A radiologist used a controlled lexicon and a tool developed for complete and standardized description of lesions to identify and annotate each lesion with semantic features. In addition, this software automatically computed image features on the basis of image texture and boundary sharpness. Semantic and computer-generated features were weighted and combined into a feature vector representing each image. An independent reference standard was created for pairwise image similarity. This was used in a leave-one-out cross-validation to train weights that optimized the rankings of images in the database in terms of similarity to query images. Performance was evaluated by using precision-recall curves and normalized discounted cumulative gain (NDCG), a common measure for the usefulness of information retrieval. RESULTS: When used individually, groups of semantic, texture, and boundary features resulted in various levels of performance in retrieving relevant lesions. However, combining all features produced the best overall results. Mean precision was greater than 90% at all values of recall, and mean, best, and worst case retrieval accuracy was greater than 95%, 100%, and greater than 78%, respectively, with NDCG. CONCLUSION: Preliminary assessment of this approach shows excellent retrieval results for three types of liver lesions visible on portal venous CT images, warranting continued development and validation in a larger and more comprehensive database.

Macrostate data clustering.

We develop an effective nonhierarchical data clustering method using an analogy to the dynamic coarse graining of a stochastic system. Analyzing the eigensystem of an interitem transition matrix identifies fuzzy clusters corresponding to the metastable macroscopic states (macrostates) of a diffusive system. A "minimum uncertainty criterion" determines the linear transformation from eigenvectors to cluster-defining window functions. Eigenspectrum gap and cluster certainty conditions identify the proper number of clusters. The physically motivated fuzzy representation and associated uncertainty analysis distinguishes macrostate clustering from spectral partitioning methods. Macrostate data clustering solves a variety of test cases that challenge other methods.

Tracing specific synonymous codon-secondary structure correlations through evolution.

We previously showed that GAU codons are preferred (relative to synonymous GAC codons) for encoding aspartates specifically at the N-termini of alpha-helices in human, but not in E. coli, proteins. To test if this difference reflected a general difference between eucaryotes and procaryotes, we now extended the analysis to include the proteins and coding sequences of mammals, vertebrates, S. cerevisiae, and plants. We found that the GAU-alpha-helix correlation is also strong in non-human mammalian and vertebrate proteins but is much weaker or insignificant in S. cerevisiae and plants. The vertebrate correlations are of sufficient strength to enhance alpha-helix N-terminus prediction. Additional results, including the observation that the correlation is significantly enhanced when proteins that are known to be correctly expressed in recombinant procaryotic systems are excluded, suggest that the correlation is induced at the level of protein translation and folding and not at the nucleic acid level. To the best of our knowledge, it is not explicable by the canonical picture of protein expression and folding, suggesting the existence of a novel evolutionary selection mechanism. One possible explanation is that some alpha-helix N-terminal GAU codons may facilitate correct co-translational folding in vertebrates.