An example bridging chemistry, biology and mental functioning
Neurotransmitters are the molecular messengers that travel across synapses between neurons, allowing communication to take place between individual cells. They consist of small chemical entities -- mostly amino acids (also the building blocks of proteins) or amines. They are released when an action potential travels down one neuron to the terminal presynaptic (transmitting) bud, and then diffuse in the intercellular fluid across the synaptic gap to the receptors on the postsynaptic (receiving) bud. Neurotransmitter uptake on the receiving end can have two different effects on the receiving neuron: increased likelihood of firing (in which case we say the neurotransmitter was excitatory), and decreased likelihood of firing (in which case we say the neurotransmitter was inhibitory).
As a use case for bridging between ontologies for mental functioning, neurobiology and chemistry, this blog post will describe the representation of neurotransmitters across bio-ontologies: chemicals in ChEBI, receptors in the NIF lexicon, various neurotransmitter-related functions and processes in GO, and mental processes and diseases involving neurotransmitter activity in MFO. Along the way, I'll make some observations about the emerging relationship between ontology and systems biology modelling.
Why are neurotransmitters important for mental functioning?
Neurotransmitters are significant chemical actors in many brain functions including emotional functioning, and deficiencies in the proper functioning of neurotransmitter pathways are implicated in various affective disorders such as bipolar disorder, addiction and depression. However, determining the exact function of a specific neurotransmitter within the overall system is difficult, since different types of neurotransmitter are often released and received in parallel, and as with all aspects of brain functioning our insights are limited by the temporal and spatial resolution we can achieve with our measurement devices. Many of our insights into the function of neurotransmitters comes from observations of the actions of drugs against dysfunctions of various sorts, where altered neurotransmitter functioning is implicated in the mechanism of action of the drug. For example, the first class of antidepressants that was discovered in the 1950's, monoamine oxidase inhibitors, act to reduce the rate at which certain sorts of neurotransmitters called catecholamines are broken down on the receiving end of the transmission, increasing the level of activity generated by the same number of neurotransmitter molecules accordingly. The antidepressant activity of these drug molecules was discovered by accident, but once this effect was found to be associated with the action of the drug on the neurotransmitter system, the role of the functioning of the neurotransmitter system in emotional well-being came to be understood. Similarly, the discovery of the antidepressant effects of serotonin reuptake inhibitors led to a greater understanding of the role of serotonin in mood. Like many other aspects of brain functioning, researchers generally agree that the exact mechanisms of action by which the molecular interventions at the subcellular level lead to the phenotypic variations at the organismal level are still not 100% understood. However, the need for further research is not a reason not to represent the current state of the knowledge we already have in ontologies. Quite the opposite: such representation might accelerate the pace at which further research is able to be conducted and integrated into the shared wealth of community knowledge. Especially when such representation is combined with predictive models that increase the capability of in silico research in the domain.
How are neurotransmitters represented in ChEBI?
ChEBI contains two primary sub-ontologies relating to molecular entities, one in which chemical entities are classified based on the sorts of chemicals they are, i.e. structural features, and another in which the ways that chemicals can act in chemical or biological systems is classified. The former is called the 'chemical entity' ontology, the latter the 'role' ontology. ‘Neurotransmitter’ is classified within the 'role' ontology as a ‘biological role’. Briefly, biological roles are the ways that chemicals can act in biological systems, and indeed this seems an appropriate classification for the job of transmitting information from one neuron to another across a synaptic gap. From an ontological perspective, we have previously argued that the biological roles of ChEBI chemicals are functions in the BFO sense. That means that they, like dispositions more generally, are the sorts of properties that represent capabilities -- things that can happen -- rather than features that are visible all the time. Functions are realized in processes. In BFO, functions are dispositions that have been selected for a particular purpose by some selection process, such as evolution in the case of biology. A particular quirk with the biological functions of small molecules is that the molecules themselves are not encoded by the genome, therefore cannot strictly be said to be the output of evolution. However, they are synthesised from dietary precursors by the biological machinery of the organism, which indeed in many cases prefers the food it does precisely because the food is abundant in the right sort of dietary precursor. Based on this chain of reasoning we can, arguably, claim that the molecule itself is also an output of a biological evolutionary selection process. But I digress.
ChEBI currently (as of the May 2012 release) contains 19 small molecules that have been assigned to the role class 'neurotransmitter'. The list includes serotonin (CHEBI:28790), acetylcholine (CHEBI:15355) and γ-aminobutyric acid (GABA, CHEBI:16865). Note that some of the types of molecule that can act as neurotransmitters can have other biological roles in addition. For example, acetylcholine can act as a vasodilator -- causing dilation of the blood vessels. Any molecule of the required type (which means having the right sort of structural composition of atoms and bonds) can act as a neurotransmitter if in the environment of the brain or as a vasodilator if in the blood -- and not every molecule of the required type will act as a neurotransmitter, since a particular molecule may never come into the neighbourhood of a neuron of that type.
How are neurotransmitters represented in Neurolex and NIFSTD?
The Neuroscience Lexicon (Neurolex) is a vocabulary for neuroscience terminology being developed in support of the data integration and semantic search interface of the NIF. It is backed by the NIFSTD ontologies, which include entities that form the core subject matter of neuroscience, such as neurons.
NeuroLex has 'neurotransmitter' in its Molecule Role ontology, which has a similar objective to the ChEBI role ontology but has specific focus on the neuroscience domain. Further information that is encoded in the NeuroLex hierarchy for neurotransmitters that is not captured in ChEBI is the distinction between inhibitory and excitatory neurotransmitter roles. NeuroLex currently lists 11 molecules as having the neurotransmitter role, and those molecules in turn are linked to ChEBI. (Yes, discussions are underway about harmonizing these two role classifications -- expect results in the near future!)
But NeuroLex has additional information beyond the 'neurotransmitter' role: it also includes the 'neurotransmitter receptor'. Interestingly, 'neurotransmitter receptor' is also a role in NeuroLex, classified beneath 'receptor role'. This highlights the fact that the same class of protein (or complex) based on their structure can serve as a neurotransmitter receptor in one environment while having alternative roles when appearing in a different environment, similarly to the case for molecules and their biological roles. NeuroLex also establishes a link between a molecule, e.g. 'serotonin', and the particular receptor, e.g. 'serotonin receptor', using the 'is related to' relationship in the underlying NIF ontology. This leaves the actual semantic specification of the nature of the relationship underspecified. One missing part of the puzzle is the observation that the disposition to act as a neurotransmitter and the disposition to act as a receptor for a neurotransmitter are mutual in the same way as a lock and key, and must be realized in the same process. Next up is a survey the relevant entities and relationships in the Gene Ontology.
Neurotransmission in the Gene Ontology
There are several neurotransmitter-related entities in the Gene Ontology. A search for 'neurotransmitter' returns a whopping 48 matches across the biological process and molecular function branches. However, many of these results are for composite processes where neurotransmission is only one component, thus are not as relevant to our discussion here. (E.g. : neurotransmitter-mediated guidance of interneurons involved in substrate-independent cerebral cortex tangential migration.) I will focus on a family of related entities: the molecular functions 'neurotransmitter binding' (GO:0042165) and 'neurotransmitter receptor activity' (GO:0030594), the biological processes 'neurotransmitter uptake' and '-secretion' (GO:0007269), some regulatory biological processes such as 'positive regulation of neurotransmitter secretion' (GO:0001956) and 'inhibition of neurotransmitter uptake' (GO:0051609), and of course the overarching 'synaptic transmission' (GO:0007268).
The way that these ontology terms relate to each other will be influenced by our ontological understanding of the sorts of entities that they are describing. The interpretation of the GO molecular functions has been contested: one possible interpretation is that they are, as the name suggests, molecular (scale) functions, thus of the same ontological kind as what are called biological roles in ChEBI and molecular roles in NeuroLex. Under this interpretation, 'neurotransmitter binding' is the disposition or function to bind to a neurotransmitter, which would inhere in any small molecule or protein that was capable of forming a chemical bond to any molecule that was capable of acting as a neurotransmitter. (Does this seem too liberal? If yes, something is implicit or missing in the definition...) 'Neurotransmitter receptor activity' is then the disposition to act as a neurotransmitter receptor, which is surely identical to the correponding 'neurotransmitter receptor' role in NeuroLex. An alternative interpretation of the GO molecular functions, suggested by the mapping to BFO discussed here and confirmed by GO editor Jane Lomax (in personal correspondence), is that they represent (perhaps small, perhaps unitary) processes. In this case, the actual process of binding -- forming a chemical bond -- is what is picked out by the ontology term 'neurotransmitter binding' and this process necessarily has at least two participants, one of which is the molecule that is capable of acting as a neurotransmitter, the other of which is the molecule capable of binding to the neurotransmitter. This is the interpretation hinted at by the provided definition "Interacting selectively and non-covalently with a neurotransmitter...". Note that in this definition, and implicitly in the term naming, the term 'neurotransmitter' is assumed to pick out a class of molecules (since a role/function itself cannot participate in a process or bind with another molecule). This presents a challenge for an increased formalization of the binding GO term, as alluded to earlier: does 'neurotransmitter binding' correctly refer to any binding of a molecule of acetylcholine (for example), even when it is acting as a vasodilator and not as a neurotransmitter? If the answer to this question is no, something is missing from our formalization, perhaps a constraint that 'neurotransmitter binding' must be realized in (if a function) or a part of (if a process) an overarching process of synaptic transmission.
The GO biological processes are more straightforwardly interpreted as processes, ontologically. For a full description of what is going on, the smaller processes 'neurotransmitter secretion' and 'neurotransmitter uptake' should be related as parts to the overarching 'synaptic transmission'. The relationship of any of these processes to the neurotransmitter molecule is one of participation, and the relationship of the process to the functions 'neurotransmitter' and 'neurotransmitter receptor' is one of realization. Both functions -- neurotransmitter and receptor -- are realized in the SAME process. There are some provisos to this simple picture, having to do with the semantics of process-process relations such as inhibition and regulation. It is important to note that the neurotransmitter function is presumably NOT realized in any process that involves 'inhibition of neurotransmitter uptake', since this is defined as 'Any process that prevents the activation of the directed movement of a neurotransmitter into a cell.'
GO is already making use of ChEBI for all references to chemical entities pariticipating in biological processes. Other projects are also using combinations of GO processes and ChEBI chemicals: for example, the Virtual Fly Brain project is backed by a fly brain anatomy ontology that defines functional categories of neurons by reference to other ontologies including GO and ChEBI. (Thanks to David Osumi-Sutherland for the details here.) One such functional category is based on the types of neurotransmitters that are released by a particular neuron type. They use a relation 'releases_neurotransmitter' that is defined as "x releases_neurotransmitter y iff:: for some 'neurotransmitter secretion ; GO:0007269' (ns), x has_function_in ns AND ns has_participant y." That means that the neuron, e.g. cholinergic neuron, has a function that is realized in a GO process of neurotransmitter secretion in which a molecule of type y, e.g. acetylcholine, participates.
Bridging biological processes and mental functioning
In a Webinar I presented recently to the NIF Webinar Series on cognitive process modelling, I mentioned that one of the objectives of the Mental Functioning Ontology project (including the Emotion Ontology specialization) is to create bridging relationships between mental processes and underlying mechanisms as represented in, for example, the GO biological processes. The example that I used was that of neurotransmitters and emotions. At present, if you want to search for all the biochemical knowledge in pathway databases such as Reactome or in modelling databases such as BioModels related to the neurobiological bases of emotions, you will struggle to find annotations that support the retrieval of pathways or models of relevance. (A search for 'emotion' in BioModels returns no results.) However, if we bridge from the Emotion Ontology e.g. 'happiness' (MFOEM:0000042), via the biological processes in GO that describe neural activity, to those neurotransmitters that are believed to play a role in mood (such as serotonin and the catecholamines), we will be able to retrieve and organise modelling results according to the mental processes -- and disorders -- we are interested in. The bridging relationship between biological processes such as 'synaptic transmission' or 'regulation of neurotransmitter uptake' to mental procesess such as 'happiness' is parthood. Less straightforward to capture is the relationship between the symptoms of diseases such as depression and mechanistic biological processes such as 'inhibition of serotonin uptake'.
Neurotransmitters are certainly not the only area where inter-ontology bridging can be used to represent important scientific knowledge about mental functioning. Hormones such as oxytocin are another very important related area -- oxytocin deficiencies (or deficiencies in pathways producing oxytocin) are implicated in various personality disorders including sociopathy. On that note, I'll leave you with this thought-provoking TED talk on a possible role for oxytocin in morality generally.