Pixel: a content management platform for quantitative omics data

Stack overview

The Pixel Web App provides researchers an intuitive way to annotate, store, explore and mine their secondary data analyses, in multi-omics biological projects. It is built upon mainstream open source technologies (see Fig. 2). Source code is hosted on the collaborative development platform GitHub (https://github.com/) and continuous integration is provided by CircleCI (https://circleci.com/). More precisely, the Pixel Web App uses the Python Django framework. This framework is based on a model-template-view architecture pattern, and data are stored in a PostgreSQL (https://www.postgresql.org/) database. We have built a docker image for the Pixel Web App. Other containers, Nginx (to serve the Django application) and PostgreSQL rely on official docker images. Each installation/deployment will result in the creation/execution of three docker instances: one for the Pixel Web App, one for the PostgreSQL database and one for the Nginx web server. In case of multiple installations, each trio of docker instances is fully isolated, meaning that data are not shared across multiple Pixel Web App installations.

Technical considerations

Installation

A step-by-step tutorial to deploy the Pixel Web App can be found in the project repository (https://github.com/Candihub/pixel/blob/master/docs-install/how-to-install.md) together with a deploy script. To summarize, this script runs the following steps:

1. Pull a tagged image of Pixel (web, see docker-composer file),

2. Start all instances (web, db and proxy) recreating the proxy and web instances. Collect all static files from the Django app. These files will be served by the proxy instance.

3. Migrate the database schema if needed (to preserve existing data).

Note that further technical considerations and full documentation can be found on GitHub repository associated to the Pixel project (https://github.com/Candihub/pixel/tree/master/docs).

Definition of terms: Omics Unit, Pixel and Pixel Set

In the Pixel Web App, the term “Omics Unit” refers to any cellular component, from any organism, which is of interest for the user. The type of Omics Unit depends on the HT experimental technology (transcriptomic, proteomic, metabolomic, etc.) from which primary and secondary datasets were collected and derived (Fig. 1A). In this context, classical Omics Units can be transcripts or proteins, but any other cellular component can be defined as, for instance, genomic regions with “peaks” in case of ChIPseq data analyses (Merhej et al., 2014). A “Pixel” refers to a quantitative measurement of a cellular activity associated to a single Omics Unit, together with a quality score (see Fig. 1A). Quantitative measurement and quality score are results of statistical analyses performed on secondary datasets, e.g., search for differentially expressed genes (Seyednasrollah, Laiho & Elo, 2015). A set of Pixels obtained from a single secondary data analysis of HT experimental results is referred as a “Pixel Set” (see Fig. 1A). Pixel Sets represent the central information in the Pixel Web App and functionalities to annotate, store, explore and mine multi-omics biological data were designed according to this concept (see below).

Functionalities to annotate, store, explore and mine Pixel Sets

Pixel Sets are obtained from secondary data analyses (see Fig. 1A). Their manipulation with the Pixel Web App consists in (i) their annotation, (ii) their storage in a database, (iii) their exploration and (iv) their mining (see Fig. 1C). This represents a cycle of multiple data analyses, which is essential in any multi-omics biological project. These different steps are detailed in the following.

Storage of Pixel Sets in the database

Import of new Pixel Sets in the Pixel Web App requires the user to follow a workflow for data submission. It corresponds to six successive steps that are explained below (Fig. 4A).

1. The “Download” step consists in downloading a template Excel file from the Pixel Web App (see Fig. 4B). In this file, multiple-choice selections are proposed for “Species”, “Strain”, “Omics Unit Type” and “Omics Area” fields. These choices reflect what is currently available in the database and can be easily expanded. User must fill other annotation fields related to the “Experiment”, “Analysis” and “Pixeler” information. The Excel file is next bundled into a ZIP archive with the secondary data file (in tab-separated values format), the user notebook (R markdown (https://rmarkdown.rstudio.com/) or Jupyter notebook (http://jupyter.org/) for instance) that contains the code used to produce the Pixel Sets from the secondary data file.

2. The “Upload” step consists in uploading the ZIP file in the Pixel Web App.

3. The step “Meta” consists in running an automatic check of the imported file integrity (md5sum checks are performed, Excel file version is verified, etc.). Note that no information is imported in the database at this stage, but a careful inspection of all Omics Units listed in the submitted Pixel Sets is done. This is why Omics Units need to be pre-registered in the Pixel Web App (see previous section).

4. In “Annotation” step, the annotations of Pixel Sets found in the Excel file (see Fig. 4C) are controlled and validated by the user.

5. Next, the “Tags” step is optional. It gives the opportunity to the user to add tags to the new Pixel Sets (see Fig. 4C), that could be helpful for further Pixel Set explorations (see next section).

6. The final step “Import archive” consists in importing all Pixel Sets in the database, together with annotations and tags.

Note that the procedure of importing metadata as an Excel file has been inspired from the import procedure widely used in GEO (Clough & Barrett, 2016).

Exploration of Pixel Sets

The Pixel Web App aims to help researchers to mine and integrate multiple Pixel Sets stored in the system. We developed a dedicated web interface to explore all the Pixel Sets stored in a particular Pixel instance (see Fig. 5).  The upper part named “Selection” lists a group of Pixel Sets selected by the user for further explorations (Fig. 5A). The middle part named “Filters” lists the Pixel database contents regarding the Species, Omics Unit Types, Omics Areas and Tags annotation fields. The user can select information (Candida glabrata and modified pH here), search and filter the Pixel Sets stored in the database (Fig. 5B). The lower part is a more flexible search field in which keywords can be type. These keywords are searched in the Analysis and Experiment detailed description fields as illustrated here with LIMMA. The web interface also comprised detailed information for the selected subset of Pixel Sets with for instance, distributions of values and quality scores and a list of individual Omics Unit shown at the bottom of the page (Fig. 5C). Note that tags have been implemented to offer to the user a versatile yet robust annotation of Pixel Sets. They are defined during the import process, but they can be modified at any time through the Pixel web interface. Once searched, matching Pixel Sets are gathered in a table that can be exported.

A case study in the pathogenic yeast Candida glabrata

The yeast Candida glabrata (C. glabrata) is a fungal pathogen of human (Bolotin-Fukuhara & Fairhead, 2014). It has been reported as the second most frequent cause of invasive infections due to Candida species, i.e., candidemia, arising especially in patients with compromised immunity (HIV virus infection, cancer treatment, organ transplantation, etc.). Candidemia remains a major cause of morbidity and mortality in the healthcare structures (Horn et al., 2009; Pfaller et al., 2012). The genome of Candida glabrata has been published in 2004 (Dujon et al., 2004). Its size is 12.3 Mb with 13 chromosomes and is composed of ∼5,200 coding regions.  Our research team is familiar with functional genomic studies in C. glabrata. In collaboration with experimental biologists, we published in the past ten years half dozen of articles, in which HT technologies were used  (Lelandais et al., 2008; Goudot et al., 2011; Merhej et al., 2015; Merhej et al., 2016; Thiébaut et al., 2017). In our lab, the Pixel Web App is installed locally and store all the necessary genomics annotations to manage any multi-omics datasets in this species.

As a case study, we decided to present how the Pixel Web App can be helpful to answer a specific biological question with only a few mouse clicks. As a biological question, we wanted to identify the genes in the entire C. glabrata genome: (i) which are annotated as involved in the yeast pathogenicity and (ii) for which the expression is significantly modified in response to an environmental stress induced by alkaline pH. Indeed, during a human host infection, C. glabrata has to face important pH fluctuations (see Ullah et al., 2013; Brunke & Hube, 2013; Linde et al., 2015 for more detailed information). Understanding the molecular processes that allow the pathogenic yeast C. glabrata to adapt extreme pH situations is therefore of medical interest to better understand host-pathogen interaction (Linde et al., 2015).

In a paper published in 2015, Linde et al. (2015) provided a detailed RNAseq based analysis of the transcriptional landscape of C. glabrata in several growth conditions, including pH shift experiments.  The primary dataset (RNAseq fastq files) is available in the Gene Expression Omnibus (Clough & Barrett, 2016) under accession number GSE61606. The secondary dataset (log2 Fold Change values) is available in Table S1 on the journal website (https://academic.oup.com/nar/article/43/3/1392/2411170). A first Pixel Set (labelled A) was created from this secondary dataset, annotated and imported into our Pixel Web App instance, following the procedure previously described. The associated ZIP archive is provided as Supplemental Information, along with all the details related to the experiment set up and the analysis. The Pixel Set A thus illustrates how publicly available data can be managed with the Pixel Web App. In our laboratory, we performed mass spectrometry experiments that also include pH shift (ZIP archive of the data is provided as Supplemental Information). Secondary dataset issued from these experiments leads to the Pixel Set B. Pixel Sets A and B comprise 5,253 Pixels and 1,879 Pixels (Fig. 6).

Transcriptomics (Pixel Set A) and proteomics (Pixel Set B) are interesting complementary multi-omics information that can be easily associated and compared with the Pixel Web App. In that respect, tags allowed to rapidly retrieve them using the web interface, applying the keywords “Candida glabrata” and “alkaline pH” (Fig. 6, Step 1). As we wanted to limit the analysis to the C. glabrata genes potentially involved in the yeast pathogenesis, a filter could be used to only retain the Omics Units for which the keyword “pathogenicity” is written in their description field (see Fig. 6, Step 2). As a result, a few numbers of Pixels were thus selected, respectively 17 in Pixel Set A and 6 in Pixel Set B. The last step consists in combining the mRNA and protein information (see Fig. 6, Step 3). For that a table comprising the multi-pixel sets can be automatically generated and easily exported. We present Table 1: five genes for which logFC values were obtained both at the mRNA and the protein levels, and for which statistical p-values were significant (<0.05). Notably two genes (CAGL0I02970g and CAGL0L08448g, lines 3 and 5 in Table 1) exhibited opposite logFC values, i.e., induction was observed at the mRNA level whereas repression was observed at the protein levels. Such observations can arise from post-translational regulation processes or from possible experimental noise, which could explain approximative mRNA or protein quantifications. In both cases, further experimental investigations are required. The three other genes (CAGL0F04807g, CAGL0F06457g and CAGL0I10516g, underlined in grey in Table 1) exhibited multi-omics coherent results and significant inductions were observed at the mRNA and protein levels. Again, further experimental investigations are required to fully validated these observations. Still, it is worth noting that the gene CAGL0F04807g, is described as “uncharacterized” in the Candida Genome Database (http://www.candidagenome.org/cgi-bin/locus.pl?locus=CAGL0F04807g&organism=C_glabrata_CBS138). Considering that logFC values for this gene are particularly high (>1), such an observation represents a good starting point to refine the functional annotation of this gene, clearly supporting the hypothesis that it has a role in the ability of C. glabrata to deal with varying pH situations.

Software availability

Pixel is released under the open-source 3-Clause BSD license (https://opensource.org/licenses/BSD-3-Clause). Its source code can be freely downloaded from the GitHub repository of the project: https://github.com/Candihub/pixel. In addition, the present version of Pixel (4.0.4) is also archived in the digital repository Zenodo (https://doi.org/10.5281/zenodo.1434316).