SUMMARY: To face up to the exponential growth of heterogeneous datasets of various organisms, we developed a user-friendly platform for building multi-omics websites, which is named Bacnet. This platform helps bioinformaticians to construct four key web interfaces: (i) an interactive genome viewer; (ii) an expression and protein atlas; (iii) an interface for analysis of co-expression network; (iv) an interface for exploring homolog presence. We believe our platform will help the bioinformaticians to construct personalized user interfaces dedicated to biologists studying non-reference organisms. AVAILABILITY AND IMPLEMENTATION: https://github.com/becavin-lab/bacnet; Java; Eclipse RAP; Eclipse RCP.

MOTIVATION: Flow and mass cytometry are experimental techniques used to measure the level of proteins expressed by cells at the single-cell resolution. Several algorithms were developed in flow cytometry to increase the number of simultaneously measurable markers. These approaches aim to combine phenotypic information of different cytometric profiles obtained from different cytometry panels. RESULTS: We present here a new algorithm, called CytoBackBone, which can merge phenotypic information from different cytometric profiles. This algorithm is based on nearest-neighbor imputation, but introduces the notion of acceptable and non-ambiguous nearest neighbors. We used mass cytometry data to illustrate the merging of cytometric profiles obtained by the CytoBackBone algorithm. AVAILABILITY AND IMPLEMENTATION: CytoBackBone is implemented in R and the source code is available at https://github.com/tchitchek-lab/CytoBackBone. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Motivation: Flow, hyperspectral and mass cytometry are experimental techniques measuring cell marker expressions at the single cell level. The recent increase of the number of markers simultaneously measurable has led to the development of new automatic gating algorithms. Especially, the SPADE algorithm has been proposed as a novel way to identify clusters of cells having similar phenotypes in high-dimensional cytometry data. While SPADE or other cell clustering algorithms are powerful approaches, complementary analysis features are needed to better characterize the identified cell clusters. Results: We have developed SPADEVizR, an R package designed for the visualization, analysis and integration of cell clustering results. The available statistical methods allow highlighting cell clusters with relevant biological behaviors or integrating them with additional biological variables. Moreover, several visualization methods are available to better characterize the cell clusters, such as volcano plots, streamgraphs, parallel coordinates, heatmaps, or distograms. SPADEVizR can also generate linear, Cox or random forest models to predict biological outcomes, based on the cell cluster abundances. Additionally, SPADEVizR has several features allowing to quantify and to visualize the quality of the cell clustering results. These analysis features are essential to better interpret the behaviors and phenotypes of the identified cell clusters. Importantly, SPADEVizR can handle clustering results from other algorithms than SPADE. Availability and Implementation: SPADEVizR is distributed under the GPL-3 license and is available at https://github.com/tchitchek-lab/SPADEVizR . Contact: [email protected]. Supplementary information: Supplementary data are available at Bioinformatics online.

MOTIVATION: Multidimensional scaling (MDS) is a well-known multivariate statistical analysis method used for dimensionality reduction and visualization of similarities and dissimilarities in multidimensional data. The advantage of MDS with respect to singular value decomposition (SVD) based methods such as principal component analysis is its superior fidelity in representing the distance between different instances specially for high-dimensional geometric objects. Here, we investigate the importance of the choice of initial conditions for MDS, and show that SVD is the best choice to initiate MDS. Furthermore, we demonstrate that the use of the first principal components of SVD to initiate the MDS algorithm is more efficient than an iteration through all the principal components. Adding stochasticity to the molecular dynamics simulations typically used for MDS of large datasets, contrary to previous suggestions, likewise does not increase accuracy. Finally, we introduce a k nearest neighbor method to analyze the local structure of the geometric objects and use it to control the quality of the dimensionality reduction. RESULTS: We demonstrate here the, to our knowledge, most efficient and accurate initialization strategy for MDS algorithms, reducing considerably computational load. SVD-based initialization renders MDS methodology much more useful in the analysis of high-dimensional data such as functional genomics datasets.