Sparse Signatures Research Articles

COADREADx: A comprehensive algorithmic dissection of colorectal cancer unravels salient biomarkers and actionable insights into its discrete progression.

Colorectal cancer is a common condition with an uncommon burden of disease, heterogeneity in manifestation, and no definitive treatment in the advanced stages. Renewed efforts to unravel the genetic drivers of colorectal cancer progression are paramount. Early-stage detection contributes to the success of cancer therapy and increases the likelihood of a favorable prognosis. Here, we have executed a comprehensive computational workflow aimed at uncovering the discrete stagewise genomic drivers of colorectal cancer progression. Using the TCGA COADREAD expression data and clinical metadata, we constructed stage-specific linear models as well as contrast models to identify stage-salient differentially expressed genes. Stage-salient differentially expressed genes with a significant monotone trend of expression across the stages were identified as progression-significant biomarkers. The stage-salient genes were benchmarked using normals-augmented dataset, and cross-referenced with existing knowledge. The candidate biomarkers were used to construct the feature space for learning an optimal model for the digital screening of early-stage colorectal cancers. The candidate biomarkers were also examined for constructing a prognostic model based on survival analysis. Among the biomarkers identified are: CRLF1, CALB2, STAC2, UCHL1, KCNG1 (stage-I salient), KLHL34, LPHN3, GREM2, ADCY5, PLAC2, DMRT3 (stage-II salient), PIGR, HABP2, SLC26A9 (stage-III salient), GABRD, DKK1, DLX3, CST6, HOTAIR (stage-IV salient), and CDH3, KRT80, AADACL2, OTOP2, FAM135B, HSP90AB1 (top linear model genes). In particular the study yielded 31 genes that are progression-significant such as ESM1, DKK1, SPDYC, IGFBP1, BIRC7, NKD1, CXCL13, VGLL1, PLAC1, SPERT, UPK2, and interestingly three members of the LY6G6 family. Significant monotonic linear model genes included HIGD1A, ACADS, PEX26, and SPIB. A feature space of just seven biomarkers, namely ESM1, DHRS7C, OTOP3, AADACL2, LPHN3, GABRD, and LPAR1, was sufficient to optimize a RandomForest model that achieved > 98% balanced accuracy (and performant recall) of cancer vs. normal on external validation. Design of an optimal multivariate model based on survival analysis yielded a prognostic panel of three stage-IV salient genes, namely HOTAIR, GABRD, and DKK1. Based on the above sparse signatures, we have developed COADREADx, a web-server for potentially assisting colorectal cancer screening and patient risk stratification. COADREADx provides uncertainty measures for its predictions and needs clinical validation. It has been deployed for experimental non-commercial use at: https://apalanialab.shinyapps.io/coadreadx/.

Compressed Sensing Σ-Δ Modulators and a Recovery Algorithm for Multi-Channel Wireless Bio-Signal Acquisition

Compressed sensing (CS) exploits signal sparsity in some domain to enable sub-Nyquist sampling which increases the energy efficiency of analog-to-digital conversion (ADC) and downstream data processing circuits–the sampling frequency is determined by the information rate, not the usual Nyquist rate. CS techniques for wireless multi-channel bio-signal recording applications based on sigma-delta modulation (SDM) are detailed and used to validate a multi-channel recovery algorithm. The SDM topology allows the required dot product calculations between the measurement and signal vectors to be performed in conjunction with its inherent integration using minimal additional circuitry. It eliminates opamp output saturation concerns and benefits directly from still ongoing Moore’s Law CMOS technology scaling. Finally, a sparse sensing matrix and recovery algorithm are described that exploit similar sparse signatures across multiple channels to improve both signal recovery accuracy and chip area efficiency. Simulation results validate the concepts.

Deep-Unfolded Sparse CDMA: Multiuser Detector and Sparse Signature Design

Sparse code division multiple access (SCDMA) is a promising non-orthogonal multiple access technique for future wireless communications. In SCDMA, transmitted symbols from multiple users are coded by their own sparse signature sequences, and a base station attempts to detect those symbols using the signature sequences. In this paper, we present a new deep-unfolded multiuser detector called a complex sparse trainable projected gradient (C-STPG) detector for SCDMA systems. Deep unfolding is a deep learning method that tunes trainable parameters in iterative algorithms using supervised data and deep learning techniques. The proposed detector provides a much superior detection performance over that of the LMMSE detector. Other advantages of the proposed detector include a low computational complexity in execution and a low training cost owing to the small number of trainable parameters. In addition, we propose a novel joint learning strategy called gradual sparsification for designing sparse signature sequences based on deep unfolding. This method is computationally efficient in optimizing a set of sparse signature sequences. Numerical results show that the gradual sparsification successfully yields sparse signature sequences with a smaller symbol error rate than those of randomly designed sparse signature sequences.

A Multistage Method for SCMA Codebook Design Based on MDS Codes

Sparse Code Multiple Access (SCMA) has been recently proposed for the future generation of wireless communication standards. SCMA system design involves specifying several parameters. In order to simplify the procedure, most works consider a multistage design approach. Two main stages are usually emphasized in these methods: sparse signatures design (equivalently, resource allocation) and codebook design. In this paper, we present a novel SCMA codebook design method. The proposed method considers SCMA codebooks structured with an underlying vector space obtained from classical block codes. In particular, when using maximum distance separable (MDS) codes, our proposed design provides maximum signal-space diversity with a relatively small alphabet. The use of small alphabets also helps to maintain desired properties in the codebooks, such as low peak-to-average power ratio and low-complexity detection.

3D seismic datasets applied to mineral exploration: revisiting three Canadian case studies

The benefits of working with legacy seismic data are broad and valued in a wide range of geoscience applications and result from the comprehensive volume of industry and public seismic data acquired over several decades. Legacy seismic data exist in many parts of the world including areas that, due to stricter regulatory processes, environmental restrictions, demographic changes, or other factors would be challenging or even impossible to survey today. Examples are numerous but one includes 2D marine seismic data acquired in the 1980s with a large airgun array to provide a deep cross-section through the Appalachians beneath the now sen-sitive Gulf of St Lawrence, eastern Canada (Marillier et al., 1989). Sometimes benefits arise through the use of newly developed tech-niques to extract additional information from legacy seismic data. This idea is obviously not new and many examples can be found in the literature. One example is the use of extended Vibroseis cor-relation developed in 1980s (Okaya and Jarchow, 1989) applied to increase the depth range of marine Vibroseis seismic data acquired in 1971 in two of the Great Lakes of North America (Milkereit et al., 1992). Marine seismic data in the Great Lakes are sparse, but contain information about deep Grenvillian terranes that, in this case, could only be revealed using a technology developed more than a decade after data acquisition. Reflection seismology has been used for mineral exploration for more than 30 years (Reed, 1993). During that period, many case studies demonstrated the utility of the method for imaging deep lithological contacts and structures indirectly associated with mineralization, and for the direct imaging of deposits. Malehmir et al. (2012) provide a comprehensive overview of key accomplishments for various commodities and mining locations in Canada and around the world. In particular, many examples demonstrate the advantages of the 3D seismic method for providing high-resolution images of subsurface structures and deep-seated ore deposits in a variety of geological environ-ments. Several of these legacy 3D seismic data sets have been reprocessed with advanced and novel techniques to provide more reliable and detailed images of the subsurface. Here, we show three early examples of the application of 3D seismic data for mineral exploration in the Bathurst, Sudbury, and Val d’Or mining camps, Canada (Figure 1). Specifically, we demonstrate that reprocessing of 3D datasets from earlier applications of seismic reflection can refine the imaging of known ore deposits and can lead to the identification of new exploration targets. For the case studies presented here, most improvements resulted from processing and imaging strategies built from an understanding of the response of mineralized bodies on seismic data and improving specific steps that prevented their proper imaging on final volumes. The case studies are introduced according to their year of acquisition, starting from the oldest (i.e., the Trill 3D survey in the Sudbury mining camp) to the most recent (e.g., the Louvicourt 3D survey in the Val d’Or mining camp). All 3D seismic surveys are located in areas with limited access and difficult terrain and as a result, were acquired with explosives. Surface conditions varied for each survey area but included steep hills, extensive swamps, and thick glacial sediments. Geology also varied at the three sites. Host rocks are either nearly-transparent-to-seismic waves or are highly heterogeneous and generating seismic scattering potentially masking the response of the mineralized body. In all three areas, however, mineralized bodies scattered or reflected energy in a preferred direction, and were observed only on a limited number of prestack seismic traces. In some cases, such a sparse signature was buried in noise when stacking all traces and required particu-lar attention to be properly imaged.

Precoding for the Sparsely Spread MC-CDMA Downlink With Discrete-Alphabet Inputs

Sparse signatures have been proposed for the CDMA uplink to reduce multi-user detection complexity, but they have not yet been fully exploited for its downlink counterpart. In this work, we propose a Multi-Carrier CDMA (MC-CDMA) downlink communication, where regular sparse signatures are deployed in the frequency domain. Taking the symbol detection point of view, we formulate a problem appropriate for the downlink with discrete alphabets as inputs. The solution to the problem provides a power-efficient precoding algorithm for the base station, subject to minimum symbol error probability (SEP) requirements at the mobile stations. In the algorithm, signature sparsity is shown to be crucial for reducing precoding complexity. Numerical results confirm system-load-dependent power reduction gain from the proposed precoding over the zero-forcing precoding and the regularized zero-forcing precoding with optimized regularization parameter under the same SEP targets. For a fixed system load, it is also demonstrated that sparse MC-CDMA with a proper choice of sparsity level attains almost the same power efficiency and link throughput as that of dense MC-CDMA yet with reduced precoding complexity, thanks to the sparse signatures.

Parsimony and goodness-of-fit in multi-dimensional NMR inversion

Multi-dimensional nuclear magnetic resonance (NMR) experiments are often used for study of molecular structure and dynamics of matter in core analysis and reservoir evaluation. Industrial applications of multi-dimensional NMR involve a high-dimensional measurement dataset with complicated correlation structure and require rapid and stable inversion algorithms from the time domain to the relaxation rate and/or diffusion domains. In practice, applying existing inverse algorithms with a large number of parameter values leads to an infinite number of solutions with a reasonable fit to the NMR data. The interpretation of such variability of multiple solutions and selection of the most appropriate solution could be a very complex problem. In most cases the characteristics of materials have sparse signatures, and investigators would like to distinguish the most significant relaxation and diffusion values of the materials.To produce an easy to interpret and unique NMR distribution with the finite number of the principal parameter values, we introduce a new method for NMR inversion. The method is constructed based on the trade-off between the conventional goodness-of-fit approach to multivariate data and the principle of parsimony guaranteeing inversion with the least number of parameter values. We suggest performing the inversion of NMR data using the forward stepwise regression selection algorithm. To account for the trade-off between goodness-of-fit and parsimony, the objective function is selected based on Akaike Information Criterion (AIC). The performance of the developed multi-dimensional NMR inversion method and its comparison with conventional methods are illustrated using real data for samples with bitumen, water and clay.

Compressed sparse time–frequency feature representation via compressive sensing and its applications in fault diagnosis

Feature extraction in time–frequency domain is wildly used in fault diagnosis of rotating machines. However, it needs more time and space to store the time–frequency information, which restricts its practical applications, especially for remote health monitoring. A novel parallel FISTA-like proximal decomposition algorithm was proposed for reconstruction of sparse time–frequency representation (TFR) from the limited noisy observations based on the recently developed compressive sensing. The effectiveness of recovering buried sparse signatures was demonstrated by numerical simulations. The proposed method yielded better results than those obtained by the traditional RecPF method. A novel framework for remote machine health condition monitoring was then developed via the proposed algorithm and the advancements in wireless communication. The effectiveness of the new proposed method for the sparse TFR in detecting bearings and gears defects in rotating machines is further verified using many practical cases. These results illustrate the proposed method can well retain TF signatures without clearly artifacts in the recovered TFR using only very limited measurements.

A New Sparse Source Separation-Based Classification Approach

In many geoscience applications, we have to convert remotely sensed images to ground cover maps. Numerous approaches to extract ground cover information have been developed. Recently, blind source separation (BSS) of remote-sensing data has received significant attention due to its suitability to recover sources when no information is available about the scanned zone, hence the term blind. In the remote-sensing context, associating each source to a significant land cover theme is difficult and constitutes the real challenge of this paper. Many authors have pointed out that BSS is overwhelmingly a question of contrast and diversity. This reasoning motivates this work which takes advantage of both decorrelation and sparsity to propose a two-level novel approach to separate our different land covers called sources. The first separation stage is based on second-order statistics or decorrelation. It gives a suitable representation of the remote-sensing images. However, decorrelation is a natural way of differentiating statistically between sources but is unable to identify and extract finer features with physical meaning. The aim of the second separation stage is to overcome this problem by an increasingly popular and powerful assumption which is the sparse representation. The last leads to good separation because most of the energy in the defined basis, at any time instant, belongs to a single source. This allows the extraction of physical features and the capture of image essential structures. The innovative aspect of this study concerns the development of a new image classification approach that integrates the BSS at the feature extraction level to provide the most relevant sources from remotely sensed images. It can be viewed as an unsupervised classification method. The second-order separation process is used as a preprocessing step to remove the interband correlation which sometimes brings ill effect to image classification. However, the second-order process is unable to uncover the underlying sources. The basic idea behind our approach is that heterogeneous multichannel data provide sparse spectral signatures in addition to sparse spatial morphologies in specified dictionaries. Hence, sparse modeling can be used to disentangle the land covers from observed mixtures. From the sparse representation, the data space is transformed into a feature space composed of mutually exclusive classes. Finally, we will merge these classes at the decision level in order to enhance the semantic capability and the reliability of land cover classification. The effectiveness of the proposed approach was demonstrated by operating two experiments to study respectively the source separation and the image classification capability of the developed approach. The different results on remote-sensing images illustrate the good performance of the new sparse approach and its robustness to noise. These experiments show that the sparse representation enhances the separation quality and allows extracting more easily the essential structures of the scanned zone. The proposed approach offers an interesting solution to the classification process with limited knowledge of ground truth.

Abstract 1998: A sparse prognostic gene signature for stage IB and II patients with completely resected non-small cell lung cancer

Abstract Background: In the statistical analysis of gene expression data, the concept sparsity of expression signature suggests that patterns of variation in expression for a larger gene set, G, could be predicted by a smaller subset of genes, g. This may be an oversimplifies the biomedical reality but a simple and meaningful model reduces the risk of overfitting and produces results which are easier to interpret. Here, we extract a sparse gene signature from the set of informative genes, and establish a functional relation between the selected genes and survival. Methods: Microarray analysis was conducted on extracted mRNA from 62 frozen JBR.10 tumor samples from patients who had undergone complete tumor resection using the Affymetrix U133A oligonucleotide microarray. We performed the following analyses on the data: (a) First, the affymetrix CEL data was processed by the BRB ArrayTools version 3.8.0. Data normalization and background correction to adjust for differences in labeling intensities was done by the GCRMA algorithm [1]. Genes showing minimal variation across arrays with expression differing by at least 1.5 fold from the median in less than 20% of the arrays were excluded from the analysis. (b) The ensuing probe sets were interrogated to identify genes whose expression was significantly related to patient survival by computing for each gene, a univariate proportional hazards models [2] evaluated at a significance threshold of 5%. (c) A sparse gene-set was then obtained from the genes correlated with survival by a multivariate Cox regression with an elastic-net [3] constraint on the regression coefficients. (d) Finally, the sparse gene-set was used to derive risk scores for all patients. Median of the risk scores was used to classify patients into low and high risk groups. Results: From the list of 8141 probe sets that passed the filtering criteria, 310 genes were significant at 0.05 level in the univariate analyses. The sparse gene-sets of 7 was obtained by applying the elastic-net variable selection and dimensionality reduction algorithm. The estimate of risk ratio associated with the high risk group versus the low risk group is 8.9 (95% CI: 3.4 − 23.5) with a p-value of 1.172*10−7 from the logrank test. Conclusions: The sparse gene-set provides an efficient set of genes that may be prognostic in patients with completely resected stage IB and Stage II non-small cell lung cancer. Validation is required in independent data sets.

Automatic generation and evaluation of sparse protein signatures for families of protein structural domains

We identified key residues from the structural alignment of families of protein domains from SCOP which we represented in the form of sparse protein signatures. A signature-generating algorithm (SigGen) was developed and used to automatically identify key residues based on several structural and sequence-based criteria. The capacity of the signatures to detect related sequences from the SWISSPROT database was assessed by receiver operator characteristic (ROC) analysis and jack-knife testing. Test signatures for families from each of the main SCOP classes are described in relation to the quality of the structural alignments, the SigGen parameters used, and their diagnostic performance. We show that automatically generated signatures are potently diagnostic for their family (ROC50 scores typically >0.8), consistently outperform random signatures, and can identify sequence relationships in the "twilight zone" of protein sequence similarity (<40%). Signatures based on 15%-30% of alignment positions occurred most frequently among the best-performing signatures. When alignment quality is poor, sparser signatures perform better, whereas signatures generated from higher-quality alignments of fewer structures require more positions to be diagnostic. Our validation of signatures from the Globin family shows that when sequences from the structural alignment are removed and new signatures generated, the omitted sequences are still detected. The positions highlighted by the signature often correspond (alignment specificity >0.7) to the key positions in the original (non-jack-knifed) alignment. We discuss potential applications of sparse signatures in sequence annotation and homology modeling.

Automated generation and refinement of protein signatures: case study with G-protein coupled receptors

Previous work had established that it was possible to derive sparse signatures (essentially sequence-length motifs) by examining points of contact between residues in proteins of known three-dimensional (3D) structure. Many interesting protein families have very little tertiary structural information. Methods for deriving signatures using only primary and secondary-structural information were therefore developed. Two methods for deriving protein signatures using protein sequence information and predicted secondary structures are described. One method is based on a scoring approach, the other on the Genetic Algorithm (GA). The effectiveness of the method was tested on the superfamily of GPCRs and compared with the established hidden Markov model (HMM) method. The signature method is shown to perform well, detecting 68% of superfamily members before the first false positive sequence and detecting several distant relationships. The GA population was used to provide information on alignment regions of particular importance for selection of key residues.

Key residues approach to the definition of protein families and analysis of sparse family signatures

Key residues approach to the definition of protein families and analysis of sparse family signatures

Alignment of a sparse protein signature with protein sequences: application to fold prediction for three small globulins

A novel algorithm has been developed for scoring the match between an imprecise sparse signature and all the protein sequences in a sequence database. The method was applied to a specific problem: signatures were derived from the probable folding nucleus and positions obtained from the determined interactions that occur during the folding of three small globular proteins and points of inter-element contact and sequence comparison of the actual three-dimensional structures of the same three proteins. In the case of two of these, lysozyme and myoglobin, the residues in the folding nucleus corresponded well to the key residues spotted by examination of the structures and in the remaining case, barnase, they did not. The diagnostic performance of the two types of signatures were compared for all three proteins. The significance of this for the application of an understanding of the protein folding mechanisms for structure prediction is discussed. The algorithm is generic and could be applied to other user-defined problems of sequence analysis.