Enhancing GO for the sake of clinical bionformatics

Anand KumarIFOMIS, University of Leipzig/Saarbrücken

Clinical Bioinformatics

Clinical Informatics / Medical Informatics Bioinformatics Biology – Medicine Bioinformatics – Medical Informatics Biology – Bioinformatics – Medical

Informatics – Medicine

Need of clinical bioinformatics

Representation of entities present at various levels of granularity

Must respect nature Must respect the professionals and the

levels of granularity they deal with Tissues : borderline?

Gene Ontology

Very widely used for annotations of gene products

Not without problems Moving target: Updates Improvements

made regularly Axes: Cellular component, Molecular

function, Biological processes

Problems with GO

Universals and particulars (Red blood cell and my Red blood cell)

Continuants and occurrants (Enzyme and enzymatic activity)

Dependent and independent entities (Respiration and Lungs)

Definitions of parents and children

Problems with GO

Parthood relations (nucleus part of cell) Circular definitions Synonyms (antigen and antibody binding) Compositional nature of terms (activator of

the establishment of competence for transformation activity)

Representation of Time

Levels of Granularity

Organism (human being) Organ System (Respiratory system,

Alimentary system) Cardinal body parts (Head, Upper limbs) Organs (Lung, Liver) Organ parts (Upper lobe of Lung, Renal

pelvis

Levels of Granularity

Tissue (Pulmonary alveolar epithelium) Tissue subdivision (Anterior epithelium of iris) Collection of cells (Menstrual secretion) Cells (Pulmonary epithelial cell) Collection of subcellular organelles (Rough

endoplasmic reticulum) Subcellular organelles (Mitochondrium) Molecules (Enzymes)

Levels of granularity

Count nouns Vs. Mass nouns Mass nouns as collections Structure Issue of Tissues

Portions? – Subdivisions Units? – Lobules not tissues Maximal portion? – Name tissue itself Parts? – Tissue subdivisions

Granularity and Parthood

Cardinal body parts and organ systems overlap.

Definitions, for examplegr(ribosome)=Subcellular

Can we call ribosome part-of hepatocyte? Parthood has a closure for both parts and

wholes

Granularity and Parthood

Not all ribosomes are present within hepatocytes Should we create a class: hepatocytic cell’s

ribosome? consider the lipopolysaccharides within the

ribosomal membrane present within the hepatocyte

Should we create a class: hepatocytic cell ribosome’s cellular membrane’s lipopolysaccharides?

Granularity and Parthood

hepatocyte*ribosome: Those ribosomes which are parts of hepatocytes

Each entity existing at a coarser level of granularity has an entity existing at each finer level of granularity as its part

Each entity existing at a finer level of granularity is a part of an entitiy existing at each coarser level of granularity

In some situations not all coarser levels exist: collection of cells, areolar tissue

Gene Ontology and Granularity

Cellular component axis contains anatomical entities cell as its highest level of granularity

Molecular functions and biological processes do not end at that level, atleast not for multicellular organisms and higher orders

Gene Ontology and Granularity

Functions and processes are dependent entities, dependent on independent entities which bear them

Since the independent entities, in this case anatomical entities, have the highest granularity as cell, any biological process which occurs at granularity coarser than cells can not get an adequate representation there

Gene Ontology and Granularity

Processes at cellular and subcellular levels could together be some or all parts of the large biological processes

Biological processes like behavior, response to extracellular stimulus, sex determination clearly have some component processes which occur at cellular level of granularity ones but not all of them

Gene Ontology and Granularity

GO has Extracellular Defined as: The space external to the outermost

structure of a cell Since there is no external limit defined, that

space could include virtually all the space within human body which are not intracellular

Not a good solution

Gene Ontology and Granularity

GO does not provide links between the three orthogonal axes, though there are various teams working on it

Problems especially when biological processes needed to be linked to cellular components

Way too many of cellular components by automatic and semi-automatic methods found eligible as bearers of the general biological processes like growth, metabolism, homeostasis

Another reason to deal with granularity

Gene Ontology and Granularity

Since GO entities are not species specific, except those with “sensu” like cytosolic ribosome (sensu Bacteria)

Difficult to understand the meaning behind entities like adult behavior

could mean a cellular level of granularity when applied to unicellular organisms and organism level of granularity for human beings

Gene Ontology and Granularity

One of the ways to get around this problem is to consider the Gene Ontology Annotations, which are species-specific

Human gene products which are annotated to entities within GO’s orthogonal axes relate those entities specific for human beings

Might mean that relations between entities within GO’s orthogonal axes would be different for different species or may not exist at all

Granularity of Functions and processes More complicated than anatomical entities Three possible ways to represent the granularity On the basis of granularity levels existing for

anatomical entities: Cellular functions (dependent on cells) exist at the cellular level of granularity. Subcellular organelle functions, organ functions and

so on this exist at the respective level of granularity of their bearers

Granularity of Functions and processes On the basis of time

Time periods can be years, months, weeks, days, hours, minutes, seconds and so on

a process which continues to take place over a longer period of time has a coarser temporal granularity as compared to processes which take place over a smaller period of time.

Granularity of Functions and processes On the basis of parthood relations existing

between various functions and processes. (Most complicated) Different functions and processes have different

number of parts and different depths to which one can identify those parts

Not all instances of a smaller process are parts of larger process (hexokinase 1 activity: glycolysis, fructose and mannose metabolism, galactose metabolism, starch and sucrose metabolism and in aminosugar metabolism )

Granularity of Functions and processes hexokinase 1 activity involved in glycolytic

pathway, to represent those cases where hexokinase 1 activity is involved in glycolysis (could be glycolytic pathway*hexokinase 1 activity)

Acknowledgements

Work on this paper was carried out under the auspices of the Wolfgang Paul Program of the Humboldt Foundation and also of the EU Network of Excellence in Semantic Datamining and the project "Forms of Life" sponsored by the Volkswagen Foundation.

Thanks to Barry Smith, Cornelius Rosse, Onard Mejino, Alan Rector, Jeremy Rogers

Enhancing GO for the sake of clinical bionformaticsAnand Kumar

IFOMIS

University of Leipzig/Saarbrücken