Towards a cellular phenotype ontologyTowards a cellular phenotype ontology

Gabriella RusticiEMBL-EBIgabry@ebi.ac.uk

• The Systems Microscopy NoE (2011 – 2015) is a FP7 funded project, involving 15 research groups across Europe

• The aim of this consortium is to develop tools and strategies to achieve a systems biology understanding of the living cell

• Done combining automated fluorescence microscopy, cell microarray, RNAi screening, quantitative image analysis and data mining

• Capture data and build models in four dimensions, three-dimensional space and time, and measure dynamic events in single living cells

• Focus on two cellular processes that are highly relevant to human cancer: cell division and cell migration

Systems Microscopy NoE at a glance

Systems Microscopy NoE at a glanceWP1

Systems Biology Analysis of Cell

Division

WP2Systems Biology Analysis of Cell

Migration

WP3Development of high throughput

imaging and screening platforms to enable production

of systems biology data

WP4Development of software for automated multi-dimensional quantitative extraction and analysis of live cell image

WP5Development of Statistics and Bioinformatics Tools for Multi-

dimensional Image-based Data

WP6Development and

Application of Modeling Methods for Systems

Microscopy

WP7Translational Applications and Outcome of Systems

Microscopy

WP8Development of Standards to Enable Systems Microscopy

WP9Development and Application

of a Public Database for Systems Microscopy

What is the current state of this field?

• Systems microscopy has not achieved the degree of standardization of other ‘omics’ approaches

• We need reporting standards, to adequately capture experimental information.

• We need a repository for quantitative data derived from images (and not a collection of raw image datasets!) that will:

1. provide access to data for the broader research community

2. accelerate the development of analytical methods for this field and

3. promote the integration of independent systems microscopy studies

What metadata/data do we need to capture?

study description including study general information and specific screen information, including protocols

siRNA library information

study results as a list of siRNAs, associated phenotypes and scores used to assign phenotypes to each siRNA

Example of study results

GENE SYMBOL ENSEMBL ID siRNA ID DEVIATIONCELLNUMBER

VALIDPLATES % INHIBITOR TRANSCRIPTS HIT TRANSCRIPT ID(S) Phenotype

ALAS1 ENSG00000023330 116885 0.750 155 4 0.50 8/8 ENST00000493402/ENSmild inhibition of secretionTOR1AIP1 ENSG00000143337 23193 0.750 102 3 0.67 7/8 ENST00000528443/ENSmild inhibition of secretionTRAPPC5 ENSG00000181029 229347 0.750 196 4 0.50 3/3 ENST00000452406/ENSmild inhibition of secretionKCTD10 ENSG00000110906 104812 0.750 87 4 0.50 3/3 ENST00000424763/ENSmild inhibition of secretionZNF252 ENSG00000196922 233977 0.750 174 4 0.75 2/4 ENST00000426361/ENSmild inhibition of secretionANXA5 ENSG00000164111 147066 0.751 272 4 0.50 4/4 ENST00000512232/ENSmild inhibition of secretionTRAM1L1 ENSG00000174599 149174 0.751 268 4 0.50 1/1 ENST00000310754 mild inhibition of secretionRSAD2 ENSG00000134321 125833 0.751 196 3 0.67 1/2 ENST00000382040 mild inhibition of secretionSMC2 ENSG00000136824 135817 0.751 271 4 0.50 5/5 ENST00000440179/ENSmild inhibition of secretionC16orf79 ENSG00000182685 215394 0.751 70 3 0.67 1/1 ENST00000328540 mild inhibition of secretionOR2T1 ENSG00000175143 237324 0.751 172 4 0.50 1/1 ENST00000366474 mild inhibition of secretionPCDHB8 ENSG00000120322 27850 0.751 343 4 0.50 1/1 ENST00000239444 mild inhibition of secretionMYL6B ENSG00000196465 11792 0.751 96 4 0.50 1/2 ENST00000207437 mild inhibition of secretionPOLG2 ENSG00000136480 120501 0.751 118 4 0.50 1/1 ENST00000322670 mild inhibition of secretionING1 ENSG00000153487 119240 0.751 178 4 0.50 3/5 ENST00000333219/ENSmild inhibition of secretionEMILIN3 ENSG00000183798 149047 0.752 297 4 0.75 1/1 ENST00000332312 mild inhibition of secretionKCTD12 ENSG00000178695 129129 0.752 277 4 0.50 2/2 ENST00000317765/ENSmild inhibition of secretionGABARAPL1 ENSG00000139112 109698 0.752 287 4 0.50 1/2 ENST00000266458 mild inhibition of secretionSQLE ENSG00000104549 118680 0.752 99 4 0.50 2/4 ENST00000265896/ENSmild inhibition of secretionPITRM1 ENSG00000107959 22207 0.752 163 4 0.50 8/11 ENST00000380994/ENSmild inhibition of secretionLMF2 ENSG00000100258 125791 0.752 181 4 0.75 4/4 ENST00000474879/ENSmild inhibition of secretionCHD3 ENSG00000170004 216241 0.752 166 4 0.50 4/8 ENST00000358181/ENSmild inhibition of secretionBAT2L1 ENSG00000130723 226775 0.752 217 4 0.50 3/9 ENST00000320547/ENSmild inhibition of secretionSLFN13 ENSG00000154760 141355 0.752 168 3 0.67 5/6 ENST00000285013/ENSmild inhibition of secretionDLX6 ENSG00000006377 224974 0.753 293 4 0.50 3/3 ENST00000437638/ENSmild inhibition of secretionSHE ENSG00000169291 228153 0.753 267 4 0.75 1/1 ENST00000304760 mild inhibition of secretionPVALB ENSG00000100362 12249 0.753 386 5 0.60 6/6 ENST00000417718/ENSmild inhibition of secretion

Search types currently supported

The current interface prototype supports 5 basic type of searches:

1. for a gene, by gene symbol or Ensembl IDs; 2. for a reagent or siRNA, by manufacturer or internal screen

ID;3. for a gene attribute, using Gene Ontology terms;4. for a phenotype, or multiple phenotypes, within an individual

screen and across screens; and5. for a study, using keywords.

Gene summary viewProvides information on a gene and the phenotypes associated with the silencing of the selected gene, across independent screens

Reagent summary viewProvides information on a siRNA reagent and the phenotypes associated with it, across independent studies

Phenotype summary view across screensProvides a list of genes, whose silencing with a specific reagents, has given rise to a particular set of phenotypes, across screens

Challenges

• Integrate data derived from independent studies and provide a meaningful representation of the experimental results

• Two levels of integration:1. At the quantitative level, through the development of

pipelines for data analysis, and2. At the level of phenotypic descriptions, through the

development of an ontology for cellular phenotypes

An ontology would help to resolve naming ambiguities (i.e. large nucleus vs large nuclei) as well as group related phenotypes together (i.e. mitotic phenotypes), facilitating the integration of independent datasets at the level of phenotypic description

Can we integrate phenotypic descriptions?

Neumann, Walter et al, 2010, Nature 464:721

Fuchs et al, 2010, Molecular Systems Biology 6: 370

Cellular Phenotype Ontology (CPO)

• Pre-composed ontology based on terms from GO BP, GO CC, GO extensions and PATO

• Split into structural (morphological) and physiological (process) abnormalities

Hoehndorf R et al. Bioinformatics 2012;28:1783-1789

Mitochondrion(GO:0005739)

Mitochondrion phenotype(CPO:XX0005739)

Mitochondrionnormal phenotype

Mitochondrionabnormal phenotype

Abnormal mitochondrion morphology Abnormal mitochondrion

physiology

Absence of mitochondrion

Cellular Phenotype Ontology (CPO)

• Physiological abnormalities are split into single and multiple occurrence processes

• PATO used to refine qualities of each

• Single occurrence processes

• Durations (increased or decreased)

• Participants (increased or decreased)

• Multiple occurrence processes

• Abnormal frequency (increased or decreased)

• Abnormal onset (increased or decreased)

Hoehndorf R et al. Bioinformatics 2012;28:1783-1789

Automatic pre-composition

• Built a beast

• 140K classes, 220K with imports

• Good underlying theoretic model

• Needs extending for Cellular Phenotype Db use case

• Potential for unrealistic classes

• EL expressivity

• Should be scalable with enough computing power

• Practically a struggle to work with on a modest PC

Entities, processes and qualities

Cellular component

Cell types

Size

Temporal quality

ShapesBiological Processes

Abnormal

Absent

Gene Ontology – Biological process

Gene Ontology – Cellular Component

Cell type ontology (CTO)

Phenotype and trait ontology (PATO)

Composing a phenotype description

• Entity Quality pattern

• Entity (a bearer of some quality)

• Quality (some characteristic of the entity)

• Phenotype: “Large nucleus”

• Entity: nucleus (GO_000xxxx)

• Quality: large (PATO_000xxxx)

• Phenotype: “Cells stuck in metaphase due to metaphase arrest”

• Entity: mitotic metaphase (GO_0000089)

• Quality: arrested (PATO_0000297)

Annotation tool

Ontology Terms

Map the phenotype terms annotated using ontologies

to CPO

Distribute tool to consortium members for phenotype

annotation

New strategy for ontology building

Phenotypes annotation tool

Original phenotypic description

Ontology based annotations

Mapping phenotypes from different biological scales

Cellular phenotypes from single cultured mammalian cells

Cellular phenotypes from mouse tissues

Cellular phenotypes from human tissues

Collect terms used to annotate cellular

phenotypes in the different domains

Map the resulting ontologies onto each other to enable correlative analysis

Open questions

• Sometimes phenotypes are not linked to biological processes because you don’t have enough information to make this association

• We mostly observe cell population phenotypes but the technology is moving towards single cell observations

• How do we deal with quantitative phenotypes (cell size, nucleus size, actin content, DNA content,…)?

• Existing ontologies might not be granular enough to describe what we want to describe

• How do we deal with the temporal information?

Goal

• Develop a data driven, generic upper level ontology for cellular phenotypes

• Open access tool for annotating phenotypes and capture necessary metadata

• Templates for new terms and ontology extension

• Pilot study with Systems Microscopy and BioMedBridgesscientists to see how they can use a EQ based annotation tool

AcknowledgementsEMBL-EBI: Catherine Kirsanova, Simon Jupp, James Malone, Alvis Brazma

EMBL-Heidelberg: Jean-Karim Heriche, Beate Neumann, Bernd Fisher, Wolfgang Huber, Jan Ellenberg, ChristophMoehl, Mayumi Isokane, Celine Revenu

CU: Robert Hoehndorf, George Gkoutos

Prototype URL: http://www.ebi.ac.uk/fg/sym