ArrayExpress and Gene Expression Atlas: Mining Functional Genomics data Gabriella Rustici, PhD Functional Genomics Team EBI-EMBL gabry@ebi.ac.uk

Talk structure Why do we need a database for functional genomics data? ArrayExpress database Archive Gene Expression Atlas Database content Query the database Data download Data submission ArrayExpress

Components of a functional genomics experiment Sample source Sample treatments RNA extraction protocol Labelling protocol Sample source Sample treatments Template preparation Array Normalized data Raw data Sample Data analysis Sample Library Chip Data analysis Library preparation Array design information Location of each element Description of each element Hybridization protocol Cluster amplification Image Scanning protocol Software specifications Sequencing and imaging Quantification matrix Software specifications From images to sequences Quality Control Sequence alignment Assembly Specific steps depending on the application Control array elements Normalization method

ArrayExpress www.ebi.ac.uk/arrayexpress/ Is a public repository for functional genomics data, mostly generated using microarray or high throughput sequencing (HTS) assays Serves the scientific community as an archive for data supporting publications, together with GEO at NCBI and CIBEX at DDBJ Provides easy access to well annotated microarray data in a structured and standardized format Facilitates the sharing of microarray designs and experimental protocols Based on FGED standards: MIAME checklist, MAGE-TAB format and MO Ontology. MINSEQE checklist for HTS data (http://www.mged.org/minseqe/) 4 ArrayExpress

Reporting standards for microarrays MIAME checklist Minimal Information About a Microarray Experiment The 6 most critical elements contributing towards MIAME are: Essential sample annotation including experimental factors and their values (e.g. compound and dose) Experimental design including sample data relationships (e.g. which raw data file relates to which sample, ….) Sufficient array annotation (e.g. gene identifiers, genomic coordinates, probe sequences or array catalog number) Essential laboratory and data processing protocols (e.g. normalization method used) Raw data for each hybridization (e.g. CEL or GPR files) Final normalized data for the set of hybridizations in the experiment 5 ArrayExpress

Reporting standards for sequencing MINSEQE checklist Minimal Information about a high-throughput Nucleotide SEQuencing Experiment The proposed guidelines for MINSEQE are (still work in progress): General information about the experiment Essential sample annotation including experimental factors and their values (e.g. compound and dose) Experimental design including sample data relationships (e.g. which raw data file relates to which sample, ….) Essential experimental and data processing protocols Sequence read data with quality scores, raw intensities and processing parameters for the instrument Final processed data for the set of assays in the experiment 6 ArrayExpress

Reporting standards for microarrays MAGE-TAB format MAGE-TAB is a simple spreadsheet format that uses a number of different files to capture information about a microarray experiment: IDF Investigation Description Format file, contains top-level information about the experiment including title, description, submitter contact details and protocols. SDRF Sample and Data Relationship Format file contains the relationships between samples and arrays, as well as sample properties and experimental factors, as provided by the data submitter. ADF Array Design Format file, describes the design of an array, i. e. the sequence located at each feature on the array and annotation of the sequences. Data files Raw and processed data files. The ‘raw’ data files are the files produced by the microarray image analysis software, such as CEL files for Affymetrix or GPR files from GenePix. The processed data file is a ‘data matrix’ file containing processed values, as provided by the data submitter. 7 ArrayExpress

Reporting standards What semantics (or ontology) should we use to best describe its annotation? Ontology, which is a formal specification of terms in a particular subject area and the relations among them. Its purpose is to provide a basic, stable and unambiguous description of such terms and relations in order to avoid improper and inconsistent use of the terminology pertaining to a given domain. Thus far, Gene Ontology (GO) has been the most successful ontology initiative. GO is a controlled vocabulary used to describe the biology of a gene product in any organism. 8 ArrayExpress

Reporting standards for microarrays MGED ontology (MO) The MO provides terms for annotating all aspects of a microarray experiment from the design of the experiment and array layout, through to the preparation of the biological sample and the protocols used to hybridize the RNA and analyze the data The MO was developed to provide terms for annotating experiments in line with the MIAME guidelines, i.e. to provide the semantics to describe a microarray experiment according to the concepts specified in MIAME Also check Open Biomedical Ontologies (OBO) initiative (www.obofoundry.org) for the development of life-science ontologies 9 ArrayExpress

ArrayExpress – two databases

How to query AE and Atlas? AE Archive Query by experiment, sample and experimental factor annotations Filter on species, array platform, molecule assayed and technology used Gene Expression Atlas Gene and/or condition queries Query across experiments and across platforms ArrayExpress

ArrayExpress – two databases 12 ArrayExpress

How much data in AE Archive? 13 ArrayExpress

Archive by species 14 ArrayExpress

Browsing the AE Archive ArrayExpress

Browsing the AE Archive The date when the data were loaded in the Archive Number of assays AE unique experiment ID Curated title of experiment Species investigated loaded in Atlas flag Raw sequencing data available in ENA The total number of experiments and assay retrieved The direct link to raw and processed data. An icon indicates that this type of data is available. The list of experiments retrieved can be printed, saved as Tab- delimited format or exported to Excel or as RSS feed 16 ArrayExpress

Browsing the AE Archive ArrayExpress

Experimental factor ontology (EFO) http://www.ebi.ac.uk/efo Application focused ontology modeling experimental factors (EFs) in AE – selected by default Developed to: increase the richness of annotations that are currently made in AE Archive to promote consistency to facilitate automatic annotation and integrate external data EFs are transformed into an ontological representation, forming classes and relationships between those classes EFO terms map to multiple existing domain specific ontologies, such as the Disease Ontology and Cell Type Ontology 18 ArrayExpress

Experimental factor ontology (EFO) An example ArrayExpress & Atlas

Searching AE Archive Simple query - EFO 20 ArrayExpress

Searching AE Archive Simple query Search across all fields: AE accession number e.g. E-MEXP-568 Secondary accession numbers e.g. GEO series accession GSE5389 Experiment name Submitter's experiment description Sample attributes, experimental factor and values, including species (e.g. GeneticModification, Mus musculus, DREB2C over-expression) Publication title, authors and journal name, PubMed ID Synonyms for terms are always included in searches e.g. 'human' and 'Homo sapiens’ 21 ArrayExpress

AE Archive query output Matches to exact terms are highlighted in yellow Matches to synonyms are highlighted in green Matches to child terms in the EFO are highlighted in pink

AE Archive – experiment view MIAME or MINSEQE scores show how much the experiment is standard compliant Link to files available. This varies between sequencing and microarray data. For microarray experiments you also have array design file and you can view a graph of the experimental design Raw data in ENA (it is a sequencing experiment), processed data downloadable as a zip Experimental factor(s) and its values 23 ArrayExpress

How does processed data look? Samples Sample annotation Genes Gene expression levels or count level data Gene annotations 24 ArrayExpress

AE Archive – SDRF file 25 ArrayExpress

SDRF file – sample & data relationship New view. Now it is easy to identify the experimental variables investigated in each experiment. In this case only ‘Developmental stage’ is the only variable, with values 10hpi, 15hpi, 20hpi, 25hpi, 30hpi, 35hpi, 40hpi, 5hpi. Also easy to see relationship between samples and data files. 26 ArrayExpress

AE Archive – ADF file 27 ArrayExpress

AE Archive – Old interface 28 ArrayExpress

AE Archive – all files 29 ArrayExpress

AE Archive – all files 30 ArrayExpress

Searching AE Archive Advanced query Combine search terms Enter two or more keywords in the search box with the operators AND, OR or NOT. AND is the default search term; a search for kidney cancer' will return hits with a match to ‘kidney' AND ‘cancer’ Search terms of more than one word must be entered inside quotes otherwise only the first word will be searched for. E.g. “kidney cancer” Specify fields for searches Particular fields for searching can also be specified in the format of fieldname:value 31 ArrayExpress

Searching AE Archive Advanced query - fieldnames Searches Example accession Experiment primary or secondary accession accession:E-MEXP-568 array Array design accession or name array:AFFY-2 OR array:Agilent* ef Experimental factor, the name of the main variables in an experiment. ef:celltype OR ef:compound efv Experimental factor value. Has EFO expansion. efv:fibroblast expdesign Experiment design type expdesign:”dose response” exptype Experiment type. Has EFO expansion. exptype:RNA-seq gxa Presence in the Gene Expression Atlas. Only value is gxa:true. ef:compound AND gxa:true pmid PubMed identifier pmid:16553887 sa Sample attribute values. Has EFO expansion. sa:wild_type species Species of the samples. Has EFO expansion. species:”homo sapiens” AND ef:cellline 32 ArrayExpress

Searching AE Archive Advanced query Filtering experiments by counts of a particular attribute Experiments fulfilling certain count criteria can also be searched for e.g. having more than 10 assays (hybridizations) Filter What is filtered assaycount:[x TO y] filter on the number of of assays where x <= y and both values are between 0 and 99,999 (inclusive) . To count excluding the values given use curly brackets e.g. assaycount:{1 TO 5} will find experiments with 2-4 assays. Single numbers may also be given e.g. assaycount:10 will find experiments with 10 assays. efcount:[x TO y] filter on the number of experimental factors samplecount:[x TO y] filter on the number of samples sacount:[x TO y] filter on the number of sample attribute categories rawcount:[x TO y] filter on the number of raw files fgemcount:[x TO y] filter on the number of final gene expression matrix (processed data) files miamescore:[x TO y] filter on the MIAME compliance score (maximum score is 5) date:yyyy-mm-dd filter by release date date:2009-12-01 - will search for experiments released on 1st of Dec 2009 date:2009* - will search for experiments released in 2009 date:[2008-01-01 2008-05-31] - will search for experiments released between 1st of Jan and end of May 2008 33 ArrayExpress

Searching AE Archive Advanced query – an example ArrayExpress & Atlas

Exercise 1 ArrayExpress

ArrayExpress – two databases 36 ArrayExpress

Gene Expression Atlas Experiment selection criteria The criteria we use for selecting experiments for inclusion in the Atlas are as follows: Array designs relating to experiment must be provided to enable re-annotation using Ensembl or Uniprot (or have the potential for this to be done) High MIAME scores Experiment must have 6 or more hybridizations Sufficient replication and large sample size EF and EFV must be well annotated Adequate sample annotation must be provided Processed data must be provided or raw data which can be renormalized must be available 37 ArrayExpress

Gene Expression Atlas Atlas construction New meta-analytical tool for searching gene expression profiles across experiments in AE Data is taken as normalized by the submitter Gene-wise linear models (limma) and t-statistics are applied to calculate the strength of genes’ differential expression across conditions across experiments The result is a two-dimensional matrix where rows correspond to genes and columns correspond to conditions, rather than samples. The matrix entries are p-values together with a sign, indicating the significance and direction of differential expression ArrayExpress

Gene Expression Atlas Atlas construction ArrayExpress

Gene Expression Atlas Atlas construction

Gene Expression Atlas ArrayExpress

Atlas home page http://www.ebi.ac.uk/gxa/ Restrict query by direction of differential expression Query for genes Query for conditions The ‘advanced query’ option allows building more complex queries ArrayExpress

Atlas home page The ‘Genes’ search box 43 ArrayExpress

Atlas home page The ‘Conditions’ search box 44 ArrayExpress

Atlas home page A single gene query 45 ArrayExpress

Atlas gene summary page 46 ArrayExpress

Atlas experiment page 47 ArrayExpress

Atlas experiment page – HTS data ArrayExpress & Atlas

Atlas home page A ‘Conditions’ only query ArrayExpress & Atlas

Atlas heatmap view ArrayExpress

Atlas list view 51 ArrayExpress Click the ‘expression profile’ link to view the experiment page 51 ArrayExpress

Atlas data download 52 ArrayExpress

Atlas gene-condition query 53 ArrayExpress

Atlas query refining 54 ArrayExpress

Atlas gene-condition query 55 ArrayExpress

Atlas query refining 56 ArrayExpress

Atlas query refining 57 ArrayExpress

Exercises 2 & 3 ArrayExpress

Data submission to AE ArrayExpress

Data submission to AE www.ebi.ac.uk/microarray/submissions.html ArrayExpress

Submission of HTS gene expression data Submit via MAGE-TAB submission route Submit: MAGE-TAB spreadsheet containing details of the samples and protocols used. Trace data files for each sample (in SRF, FASTQ or SFF format ) Processed data files  For non-human species we will supply your SRF or FASTQ files to the European Nucleotide Archive (ENA). If you have human identifiable sequencing data you need to submit to the The European Genome-phenome Archive and not ArrayExpress. They will supply you with a suitable template for submission and store human identifiable data securely. ArrayExpress & Atlas

Types of data that can be submitted ArrayExpress & Atlas

What happens after submission? Email confirmation Curation The curation team will review your submission and will email you with any questions. Possible reopening for editing We will send you an accession number when all the required information has been provided. We will load your experiment into ArrayExpress and provide you with a reviewer login for viewing the data before it is made public. ArrayExpress & Atlas