Unsupervised classification to improve the quality of a bird song recording dataset

Open audio databases such as Xeno-Canto are widely used to build datasets to explore bird song repertoire or to train models for automatic bird sound classification by deep learning algorithms. However, such databases suffer from the fact that bird sounds are weakly labelled: a species name is attributed to each audio recording without timestamps that provide the temporal localization of the bird song of interest. Manual annotations can solve this issue, but they are time consuming, expert-dependent, and cannot run on large datasets. Another solution consists in using a labelling function that automatically segments audio recordings before assigning a label to each segmented audio sample. Although labelling functions were introduced to expedite strong label assignment, their classification performance remains mostly unknown. To address this issue and reduce label noise (wrong label assignment) in large bird song datasets, we introduce a data-centric novel labelling function composed of three successive steps: 1) time-frequency sound unit segmentation, 2) feature computation for each sound unit, and 3) classification of each sound unit as bird song or noise with either an unsupervised DBSCAN algorithm or the supervised BirdNET neural network. The labelling function was optimized, validated, and tested on the songs of 44 West-Palearctic common bird species. We first showed that the segmentation of bird songs alone aggregated from 10% to 83% of label noise depending on the species. We also demonstrated that our labelling function was able to significantly reduce the initial label noise present in the dataset by up to a factor of three. Finally, we discuss different opportunities to design suitable labelling functions to build high-quality animal vocalizations with minimum expert annotation effort.     

Classification of Time Series Gene Expression in Clinical Studies via Integration of Biological Network

The increasing availability of time series expression datasets, although promising, raises a number of new computational challenges. Accordingly, the development of suitable classification methods to make reliable and sound predictions is becoming a pressing issue. We propose, here, a new method to classify time series gene expression via integration of biological networks. We evaluated our approach on 2 different datasets and showed that the use of a hidden Markov model/Gaussian mixture models hybrid explores the time-dependence of the expression data, thereby leading to better prediction results. We demonstrated that the biclustering procedure identifies function-related genes as a whole, giving rise to high accordance in prognosis prediction across independent time series datasets. In addition, we showed that integration of biological networks into our method significantly improves prediction performance. Moreover, we compared our approach with several state-of–the-art algorithms and found that our method outperformed previous approaches with regard to various criteria. Finally, our approach achieved better prediction results on early-stage data, implying the potential of our method for practical prediction.     

JSOM: Jointly-evolving self-organizing maps for alignment of biological datasets and identification of related clusters

With the rapid advances of various single-cell technologies, an increasing number of single-cell datasets are being generated, and the computational tools for aligning the datasets which make subsequent integration or meta-analysis possible have become critical. Typically, single-cell datasets from different technologies cannot be directly combined or concatenated, due to the innate difference in the data, such as the number of measured parameters and the distributions. Even datasets generated by the same technology are often affected by the batch effect. A computational approach for aligning different datasets and hence identifying related clusters will be useful for data integration and interpretation in large scale single-cell experiments. Our proposed algorithm called JSOM, a variation of the Self-organizing map, aligns two related datasets that contain similar clusters, by constructing two maps—low-dimensional discretized representation of datasets–that jointly evolve according to both datasets. Here we applied the JSOM algorithm to flow cytometry, mass cytometry, and single-cell RNA sequencing datasets. The resulting JSOM maps not only align the related clusters in the two datasets but also preserve the topology of the datasets so that the maps could be used for further analysis, such as clustering.     

Application of a self-organizing map to select representative species in multivariate analysis: A case study determining diatom distribution patterns across France

Ecological communities consist of a large number of species. Most species are rare or have low abundance, and only a few are abundant and/or frequent. In quantitative community analysis, abundant species are commonly used to interpret patterns of habitat disturbance or ecosystem degradation. Rare species cause many difficulties in quantitative analysis by introducing noises and bulking datasets, which is worsened by the fact that large datasets suffer from difficulties of data handling. In this study we propose a method to reduce the size of large datasets by selecting the most ecologically representative species using a self organizing map (SOM) and structuring index (SI). As an example, we used diatom community data sampled at 836 sites with 941 species throughout the French hydrosystem. Out of the 941 species, 353 were selected. The selected dataset was effectively classified according to the similarities of community assemblages in the SOM map. Compared to the SOM map generated with the original dataset, the community pattern gave a very similar representation of ecological conditions of the sampling sites, displaying clear gradients of environmental factors between different clusters. Our results showed that this computational technique can be applied to preprocessing data in multivariate analysis. It could be useful for ecosystem assessment and management, helping to reduce both the list of species for identification and the size of datasets to be processed for diagnosing the ecological status of water courses.     

An optimized data structure for high-throughput 3D proteomics data: mzRTree

As an emerging field, MS-based proteomics still requires software tools for efficiently storing and accessing experimental data. In this work, we focus on the management of LC–MS data, which are typically made available in standard XML-based portable formats. The structures that are currently employed to manage these data can be highly inefficient, especially when dealing with high-throughput profile data. LC–MS datasets are usually accessed through 2D range queries. Optimizing this type of operation could dramatically reduce the complexity of data analysis. We propose a novel data structure for LC–MS datasets, called mzRTree, which embodies a scalable index based on the R-tree data structure. mzRTree can be efficiently created from the XML-based data formats and it is suitable for handling very large datasets. We experimentally show that, on all range queries, mzRTree outperforms other known structures used for LC–MS data, even on those queries these structures are optimized for. Besides, mzRTree is also more space efficient. As a result, mzRTree reduces data analysis computational costs for very large profile datasets.     

A comparison of supermatrix and supertree methods for multilocus phylogenetics using organismal datasets

It has been proposed that supertree approaches should be applied to large multilocus datasets to achieve computational tractability. Large datasets such as those derived from phylogenomics studies can be broken into many locus‐specific tree searches and the resulting trees can be stitched together via a supertree method. Using simulated data, workers have reported that they can rapidly construct a supertree that is comparable to the results of heuristic tree search on the entire dataset. To test this assertion with organismal data, we compare tree length under the parsimony criterion and computational time for 20 multilocus datasets using supertree (SuperFine and SuperTriplets) and supermatrix (heuristic search in TNT) approaches. Tree length and computational times were compared among methods using the Wilcoxon matched‐pairs signed rank test. Supermatrix searches produced significantly shorter trees than either supertree approach (SuperFine or SuperTriplets; P < 0.0002 in both cases). Moreover, the processing time of supermatrix search was significantly lower than SuperFine+locus‐specific search (P < 0.01) but roughly equivalent to that of SuperTriplets+locus‐specific search (P > 0.4, not significant). In conclusion, we show by using real rather than simulated data that there is no basis, either in time tractability or in tree length, for use of supertrees over heuristic tree search using a supermatrix for phylogenomics.     

Computational methods for ultrastructural analysis of synaptic complexes

Electron microscopy (EM) provided fundamental insights about the ultrastructure of neuronal synapses. The large amount of information present in the contemporary EM datasets precludes a thorough assessment by visual inspection alone, thus requiring computational methods for the analysis of the data. Here, I review image processing software methods ranging from membrane tracing in large volume datasets to high resolution structures of synaptic complexes. Particular attention is payed to molecular level analysis provided by recent cryo-electron microscopy and tomography methods.     

Effective and scalable single-cell data alignment with non-linear canonical correlation analysis

Data alignment is one of the first key steps in single cell analysis for integrating multiple datasets and performing joint analysis across studies. Data alignment is challenging in extremely large datasets, however, as the major of the current single cell data alignment methods are not computationally efficient. Here, we present VIPCCA, a computational framework based on non-linear canonical correlation analysis for effective and scalable single cell data alignment. VIPCCA leverages both deep learning for effective single cell data modeling and variational inference for scalable computation, thus enabling powerful data alignment across multiple samples, multiple data platforms, and multiple data types. VIPCCA is accurate for a range of alignment tasks including alignment between single cell RNAseq and ATACseq datasets and can easily accommodate millions of cells, thereby providing researchers unique opportunities to tackle challenges emerging from large-scale single-cell atlas.     

HaploBlocks: Efficient Detection of Positive Selection in Large Population Genomic Datasets

Genomic regions under positive selection harbor variation linked for example to adaptation. Most tools for detecting positively selected variants have computational resource requirements rendering them impractical on population genomic datasets with hundreds of thousands of individuals or more. We have developed and implemented an efficient haplotype-based approach able to scan large datasets and accurately detect positive selection. We achieve this by combining a pattern matching approach based on the positional Burrows–Wheeler transform with model-based inference which only requires the evaluation of closed-form expressions. We evaluate our approach with simulations, and find it to be both sensitive and specific. The computational resource requirements quantified using UK Biobank data indicate that our implementation is scalable to population genomic datasets with millions of individuals. Our approach may serve as an algorithmic blueprint for the era of “big data” genomics: a combinatorial core coupled with statistical inference in closed form.     

Two-stage clustering (TSC): a pipeline for selecting operational taxonomic units for the high-throughput sequencing of PCR amplicons

Clustering 16S/18S rRNA amplicon sequences into operational taxonomic units (OTUs) is a critical step for the bioinformatic analysis of microbial diversity. Here, we report a pipeline for selecting OTUs with a relatively low computational demand and a high degree of accuracy. This pipeline is referred to as two-stage clustering (TSC) because it divides tags into two groups according to their abundance and clusters them sequentially. The more abundant group is clustered using a hierarchical algorithm similar to that in ESPRIT, which has a high degree of accuracy but is computationally costly for large datasets. The rarer group, which includes the majority of tags, is then heuristically clustered to improve efficiency. To further improve the computational efficiency and accuracy, two preclustering steps are implemented. To maintain clustering accuracy, all tags are grouped into an OTU depending on their pairwise Needleman-Wunsch distance. This method not only improved the computational efficiency but also mitigated the spurious OTU estimation from 'noise' sequences. In addition, OTUs clustered using TSC showed comparable or improved performance in beta-diversity comparisons compared to existing OTU selection methods. This study suggests that the distribution of sequencing datasets is a useful property for improving the computational efficiency and increasing the clustering accuracy of the high-throughput sequencing of PCR amplicons. The software and user guide are freely available at http://hwzhoulab.smu.edu.cn/paperdata/.     

Knime4Bio: a set of custom nodes for the interpretation of next-generation sequencing data with KNIME

SUMMARY: Analysing large amounts of data generated by next-generation sequencing (NGS) technologies is difficult for researchers or clinicians without computational skills. They are often compelled to delegate this task to computer biologists working with command line utilities. The availability of easy-to-use tools will become essential with the generalization of NGS in research and diagnosis. It will enable investigators to handle much more of the analysis. Here, we describe Knime4Bio, a set of custom nodes for the KNIME (The Konstanz Information Miner) interactive graphical workbench, for the interpretation of large biological datasets. We demonstrate that this tool can be utilized to quickly retrieve previously published scientific findings.     

Analyzing large datasets with bootstrap penalization

Data with a large p (number of covariates) and/or a large n (sample size) are now commonly encountered. For many problems, regularization especially penalization is adopted for estimation and variable selection. The straightforward application of penalization to large datasets demands a “big computer” with high computational power. To improve computational feasibility, we develop bootstrap penalization, which dissects a big penalized estimation into a set of small ones, which can be executed in a highly parallel manner and each only demands a “small computer”. The proposed approach takes different strategies for data with different characteristics. For data with a large p but a small to moderate n, covariates are first clustered into relatively homogeneous blocks. The proposed approach consists of two sequential steps. In each step and for each bootstrap sample, we select blocks of covariates and run penalization. The results from multiple bootstrap samples are pooled to generate the final estimate. For data with a large n but a small to moderate p, we bootstrap a small number of subjects, apply penalized estimation, and then conduct a weighted average over multiple bootstrap samples. For data with a large p and a large n, the natural marriage of the previous two methods is applied. Numerical studies, including simulations and data analysis, show that the proposed approach has computational and numerical advantages over the straightforward application of penalization. An R package has been developed to implement the proposed methods.     

scAB detects multiresolution cell states with clinical significance by integrating single-cell genomics and bulk sequencing data

Although single-cell sequencing has provided a powerful tool to deconvolute cellular heterogeneity of diseases like cancer, extrapolating clinical significance or identifying clinically-relevant cells remains challenging. Here, we propose a novel computational method scAB, which integrates single-cell genomics data with clinically annotated bulk sequencing data via a knowledge- and graph-guided matrix factorization model. Once combined, scAB provides a coarse- and fine-grain multiresolution perspective of phenotype-associated cell states and prognostic signatures previously not visible by single-cell genomics. We use scAB to enhance live cancer single-cell RNA-seq data, identifying clinically-relevant previously unrecognized cancer and stromal cell subsets whose signatures show a stronger poor-survival association. The identified fine-grain cell subsets are associated with distinct cancer hallmarks and prognosis power. Furthermore, scAB demonstrates its utility as a biomarker identification tool, with the ability to predict immunotherapy, drug responses and survival when applied to melanoma single-cell RNA-seq datasets and glioma single-cell ATAC-seq datasets. Across multiple single-cell and bulk datasets from different cancer types, we also demonstrate the superior performance of scAB in generating prognosis signatures and survival predictions over existing models. Overall, scAB provides an efficient tool for prioritizing clinically-relevant cell subsets and predictive signatures, utilizing large publicly available databases to improve prognosis and treatments.     

Multiple protein sequence alignment

Multiple sequence alignments are essential in computational analysis of protein sequences and structures, with applications in structure modeling, functional site prediction, phylogenetic analysis and sequence database searching. Constructing accurate multiple alignments for divergent protein sequences remains a difficult computational task, and alignment speed becomes an issue for large sequence datasets. Here, I review methodologies and recent advances in the multiple protein sequence alignment field, with emphasis on the use of additional sequence and structural information to improve alignment quality.     

Physical properties of biological entities: an introduction to the ontology of physics for biology

As biomedical investigators strive to integrate data and analyses across spatiotemporal scales and biomedical domains, they have recognized the benefits of formalizing languages and terminologies via computational ontologies. Although ontologies for biological entities-molecules, cells, organs-are well-established, there are no principled ontologies of physical properties-energies, volumes, flow rates-of those entities. In this paper, we introduce the Ontology of Physics for Biology (OPB), a reference ontology of classical physics designed for annotating biophysical content of growing repositories of biomedical datasets and analytical models. The OPB's semantic framework, traceable to James Clerk Maxwell, encompasses modern theories of system dynamics and thermodynamics, and is implemented as a computational ontology that references available upper ontologies. In this paper we focus on the OPB classes that are designed for annotating physical properties encoded in biomedical datasets and computational models, and we discuss how the OPB framework will facilitate biomedical knowledge integration.     

AutoSOME: a clustering method for identifying gene expression modules without prior knowledge of cluster number

Background  

Clustering the information content of large high-dimensional gene expression datasets has widespread application in "omics" biology. Unfortunately, the underlying structure of these natural datasets is often fuzzy, and the computational identification of data clusters generally requires knowledge about cluster number and geometry.     

Combined dynamic arrays for storing and searching semi-ordered tandem mass spectrometry data.

When performing bioinformatics analysis on tandem mass spectrometry data, there is a computational need to efficiently store and sort these semi-ordered datasets. To solve this problem, a new data structure based on dynamic arrays was designed and implemented in an algorithm that parses semi-ordered data made by Mascot, a separate software program that matches peptide tandem mass spectra to protein sequences in a database. By accommodating the special features of these large datasets, the combined dynamic array (CDA) provides efficient searching and insertion operations. The operations on real datasets using this new data structure are hundreds times faster than operations using binary tree and red-black tree structures. The difference becomes more significant when the dataset size grows. This data structure may be useful for improving the speed of other related types of protein assembling software or other types of software that operate on datasets with similar semi-ordered features.     

Statistical methods and software for the analysis of highthroughput reverse genetic assays using flow cytometry readouts

Highthroughput cell-based assays with flow cytometric readout provide a powerful technique for identifying components of biologic pathways and their interactors. Interpretation of these large datasets requires effective computational methods. We present a new approach that includes data pre-processing, visualization, quality assessment, and statistical inference. The software is freely available in the Bioconductor package prada. The method permits analysis of large screens to detect the effects of molecular interventions in cellular systems.     

Hierarchical spatial modeling of additive and dominance genetic variance for large spatial trial datasets

Summary .  This article expands upon recent interest in Bayesian hierarchical models in quantitative genetics by developing spatial process models for inference on additive and dominance genetic variance within the context of large spatially referenced trial datasets. Direct application of such models to large spatial datasets are, however, computationally infeasible because of cubic-order matrix algorithms involved in estimation. The situation is even worse in Markov chain Monte Carlo (MCMC) contexts where such computations are performed for several iterations. Here, we discuss approaches that help obviate these hurdles without sacrificing the richness in modeling. For genetic effects, we demonstrate how an initial spectral decomposition of the relationship matrices negate the expensive matrix inversions required in previously proposed MCMC methods. For spatial effects, we outline two approaches for circumventing the prohibitively expensive matrix decompositions: the first leverages analytical results from Ornstein–Uhlenbeck processes that yield computationally efficient tridiagonal structures, whereas the second derives a modified predictive process model from the original model by projecting its realizations to a lower-dimensional subspace, thereby reducing the computational burden. We illustrate the proposed methods using a synthetic dataset with additive, dominance, genetic effects and anisotropic spatial residuals, and a large dataset from a Scots pine ( Pinus sylvestris L.) progeny study conducted in northern Sweden. Our approaches enable us to provide a comprehensive analysis of this large trial, which amply demonstrates that, in addition to violating basic assumptions of the linear model, ignoring spatial effects can result in downwardly biased measures of heritability.