370 likes | 458 Views
An evolutionary approach to function (and roles!). Phillip Lord Newcastle University Bio-Ontologies 2009 http://hdl.handle.net/10101/npre.2009.3228.1. Overview. Existing definitions of function, role and biological function are limited
E N D
An evolutionary approach to function (and roles!) Phillip Lord Newcastle University Bio-Ontologies 2009 http://hdl.handle.net/10101/npre.2009.3228.1
Overview • Existing definitions of function, role and biological function are limited • I provide alternative definitions to overcome these limitations • The definitions seek not to change current usage but formalize it.
Introduction • Lots of biology consists of analysing function • The heart pumps blood in circles • The brain is the bit we think with • (sometimes) • Subcellular homeostasis of phytohormone auxin is mediated by the ER-localized PIN5 transporter doi:10.1038/nature08066
Introduction • We've ontologized it. • But what actually is it!
Methodology • Why do we need function in an ontology • What are the current definitions • Evolve new definitions from examples • Test these against a substantial independent usage. • Aim to give definitions which are simple, clear and, critically, applicable.
Why do we need function • Phil participates_in Talking • A thing (that's me) is involved in a process (that's this). • Why is this not enough? • Phil is able to participate in Talking • Phil is paid to participate in Talking • Phil is meant to participate in Talking • These relationships are prime candidates for an upper ontology
Modelling Function Phil Phil bearer_of Disposition participates_in Function Role realized_in Talking Talking
BFO's treatment • BFO – Basic Formal Ontology • Small upper ontology meant to provide key concepts for use in biomedicine. • Used within some OBOFoundary ontologies • Includes concepts for function
BFO:Function Function is a realizable entity the manifestation of which is an essentially end-directed activity of a continuant entity being a specific kind of entity in the kind or kinds of context that it is made for. In English: A hammer was made to hammer nails in a hammering process
But what about biology? • The previous definition works as a definition of ArtifactualFunction • An organism isn't made for anything. So, no functions. • So, we need a definition for biological function
BFO:Biological Function A biological function is a function which inheres in an independent continuant that is i) part of an organism and ii) exists and has the physical structure as a result of the coordinated expression of that organism's structural genes. http://hdl.handle.net/10101/npre.2008.1941.1 In English: A foot is part of an organism, exists and develops in a coordinated way and we walk on it.
Biological Function • But: • A differentiated tumour • Is part of organism • Grows • So can have a function • A male ant • Is not part of an organism • So cannot have a function • it appears to have genetic eusocial function • A protein • Has a structure which does not depend on structural genes • So cannot have a function
My new definition A biological function is a realizable entity where the homologous structure(s) of individuals of closely related (or the same) species bear this same biological function. • In English: My feet function in walking because I walk on them, so do you, and so does my pet chimp • A tumour has no homologs • So has no function • A male ant has homologs and they behave in the same way • So can have a function • Likewise, proteins
How do the functions relate? Sole of Foot realized_in is_bearer Shock Resistance Bio. Function • What distinguishes a biological function from an artifactual function? • The process? Shock Resistance? • But the sole of my shoe does the same • But the sole of my shoe is not an organism or part of one • It is the bearer which makes the difference, not the process
How do the functions relate • Can a function be both a biological and artifactual function? • A bacteria produced using synthetic biology techniques which changes colour in the presence of a toxin. • So, probably we can be both at the same time.
So, a definition of function • A function is a realizable entity which is either a biological function or an artifactual function.
Roles • A front foot has function to resist shock in a process of shock resistance • But humans do not walk on their hands, although they can, mostly for fun, or for show • So, the ability to resist shock "can be served or participated in by that kind of continuant in some kinds of natural, social or institutional context" • This is a BFO:Role • A hand has a role to resist shock in a process of shock resistance
Roles • So, it is the manner of the bearing relationship which makes the distinction • A concept like "to resist shock", can have instances which are roles, biological or artifactual functions. • My model is descriptive of my examples, but is it predictive: does it work against usage from other people?
In practice with OBI • Ontology for Biomedical Investigations • Describes the equipment, people and technology behind an experiment • OBI was built post-BFO and with knowledge of it • Has many concepts under function and role.
OBI Functions • Perturb • I can perturb a tube with my fingers • Measure • I can measure a distance with my arm • Most OBI Functions are not necessarily functions • Full details in paper
OBI Roles • Label Role(reagent role realized in detection of label assay) • But S35 CTP was produced specifically for this purpose • Reference Role(support observation of relative magnitude) • The international kilogram prototype was manufactured for this purpose http://en.wikipedia.org/wiki/File:Silicon_sphere_for_Avogadro_project.jpg http://en.wikipedia.org/wiki/File:CGKilogram.jpg
Summary so far • I define biological function by homology • Artifactual and Biological Function are probably not disjoint • I define Function as one or the other • Many entities can be either roles or functions (or subcategories)
Axiomatisation • A relationship between RealizableEntity and Process • Function and Role are defined subclasses • Most children asserted under RealizableEntity, • Individuals inferred under Function (bio or artifactual) or Role
How to apply these definitions • How do I know if I have a function (bio or arti…) or role? • What is my assay? • Artifactual Function • Was the artifact intended for a purpose? • We can guess this for ancient artifacts. • Does the structure support this purpose?
Applying biological function • Biological Function • Measures exist for "closely related" and for "homolog". • Other definitions are less applicable • Definitions based on "caused by evolution" are imprecise – which bit of evolution? • Definitions based on selection are: • Hard or impossible to measure • Natural, sexual or artificial? • Ignore drift, founder effects or other forms of neutral evolution.
Summary • Existing definitions of function, role and biological function are limited • These limitations have been demonstrated using OBI • I provide alternative definitions to overcome this. • The definitions seek not to change current usage but formalize it.
Acknowledgements • From NCL • Frank Gibson (now abcam) • Matt Pocock • Dan Swan • Simon Cockell • James Malone, Helen Parkinson, Bjoern Peters and all at OBI • Robert Stevens • The reviewers!
Questions • Does BFO2.0 not fix this? • What about organisms with no homologs? • What about "gain-of-function" mutations? • What about hyper variable genes? • What about chemical function? • Does this work with GO? • You used OWL, what about OBO? • Why homology, rather than orthology? • But the Kilogram prototype will lose it's role soon? • How will this help? Is it the cure for old age?
Does BFO 2.0 not fix this • No. BFO 2.0 uses an "evolved this way" clause which I think is problematic. • It also uses the notion of a typical life plan which is hard to measure.
What about organisms with no homologs? • Surely, if there are no homologs there are no functions? • First, all organisms have homologs – the definition does not require that they be extant. • Not all structures do. See "what about gain-of-function" mutations. • If we can't find the homologs, then it means we don't know whether we have a function, not that there isn't one.
What about "gain-of-function" mutations? • What about hyper variable genes? • A gain-of-function mutation, as a new mutation has no homolog. We would therefore use a role or a realizable entity. • Also, for hyper-variable genes. • At heart, I side with Dumontier(2008) • Genes and proteins only do what their structure allows. • And they are capable of doing anything that their structure allows. • The distinctions we make between roles and functions are just not that useful at this level of granularity. • My definition is at least applicable. Whether the distinction is useful is a different question.
What about chemical function? • We may need a definition for chemical function, but I am not sure; I don't have the examples yet. • As with genes and proteins, I am not that convinced that the distinction between role and function is that useful at this level. • See "gain-of-function".
Does this work with GO? • Yes, as far as I can tell, although I have not done a complete survey. • Homology is already used to determine function. • In most cases, the evidence codes tell us whether we have enough information to be sure we have a function as opposed to realizable entity. • Molecular function would actually become a child of realisable entity – the naming is unwieldy, but would have little other consequences. This is also the case with BFO now as it stands. • I'm not convinced the distinction between role and function is that useful for genes. See "gain-of-function"
You used OWL, what about OBO? • I believe that my axiomatisation is mostly translatable into OBO • I think OBO lacks an universal quantification relationship which I used. • The English definitions are useful irrespective of the axiomatisation.
Why homology rather than orthology? • In some ways orthology is a more natural fit, to exclude paralogy. • But the distinction between orthology and paralogy is normally made only at a genetic level; so the more general term makes sense • Especially given that the role/function distinction is least useful at a genetic level. • See "gain-of-function"
But the Kilogram prototype will lose it's role soon? • The international kilo will no longer be a reference soon. The prototype never changed. So it's a role? • Perhaps. But it's highly manufactured state suggests otherwise. • My imperial spanner set has no function now, because I can't buy nuts and bolts in inches. But the spanners have not changed. Is this a role too?
How will this help? Is it the cure for old age? • It helps insofar as an upper ontology helps • My experience suggests upper ontologies can help in building new ontologies • But only if its obvious how to apply the definitions outside of their original use • It may help interoperability between ontologies • Although, it's not clear how often we care about these top level distinctions when querying • My definitions do avoid problems of scale in biology. • While the definitions work, it's not clear that the distinctions are useful at all levels of granularity. • Yes, it will cure old age. • This is a joke