Schizophrenia and bipolar disorder genetics blog

While all would agree that the belief of the layman (read media) that ’schizophrenia’ refers to a split mind or split personality is highly irritating, I think the views expressed by these people are unhelpful.

This action group known as CASL aim to do away with the term schizophrenia but do not specify what it should be replaced with.

As a word, schizophrenia doesn’t seem to have that vaguely pejoritive feel that former medical terms for learning disability/mental retardation did. I can understand those people with the disorder not wishing to be known as ’schizophrenic’ but rather ‘diagnosed with schizophrenia’…..this latter phrasing maintains their human individuality.

My hope is that the fields of genetics and molecular neurobiology will be able to provide a more fact-based subdivision of psychiatric illnesses into particular categories. Until that time, I think it is premature to be thinking about new names. That would only act to distract from the real issues of treatment and scientific investigation.

The last few months have been a good time for readers of reviews in the field of schizophrenia and bipolar disorder genetics. What makes the recent batch more impressive is that they are not merely collating primary research but using that data to construct theories and models that seek to provide an overview of the biological systems failing in these conditions. It is that ‘added value’ that makes these papers ‘even better than the real thing’, where the ‘real thing’ is the incremental output of most primary papers. Below I’ve listed three of the best and provided a summary for those who can’t access the primary papers.

A pathological paradigm for the therapy of psychiatric disease
Spedding, Jay, Costa e Silva and Perret.

Not a light read but an important review nevertheless. This paper deals with the past, present and future of drug therapies for psychiatric illness, perhaps with a bias towards depression and affective disorders, but with valuable lessons for schizophrenia and bipolar disorder too. What separates this from just a summary of drug trials and clinical outcomes is that the authors have a set of criteria they believe will lead to better therapies. These are: an understanding of the precise brain regions affected in each disorder and the targeting of specific drugs to them and, once there, these drugs should re-establish normal physiological ‘set points’. According to the authors, these set points can be thought of as changeable states of ‘plasticity’ - the levels of connectedness/communication between nerve cells.
Like the next review, the authors have focussed on a few key neural circuits in which plasticity properties have been well described. They integrate this knowledge with our current understanding of several polymorphisms in neurotransmitter systems as well as highlighting the influence of stress on these pathways.
The authors have, therefore, managed to bridge the gaps between several disparate fields of research in an attempt to synthesize a holistic model for rational therapy design. While not fully explaining how measurable readouts for therapy-mediated changes in plasticity might be obtained, the review does a good job in exposing the reader to the various levels of physiology affected by these conditions.

Intermediate phenotypes and genetic mechanisms of psychiatric disorders
Meyer-Lindenberg and Weinberger

What is an ‘intermediate phenotype’? To quote the authors:

“As genes do not encode for psychopathology, it is reasonable to expect that the association or penetrance of gene effects will be greater at the level of relatively more simple and biologically based phenotypes.�

Put simply, a car is not made for breaking down (except a white Ford Fiesta ‘Popular Plus’ 1986 vintage which I spent ££££ on but sold for enough money to buy a tub of ice cream)…. it’s made for travelling. Therefore, it is not always immediately obvious from the outside what is going on when it stops working. However, if you have a rudimentary understanding of the normal performance of basic engine systems then you might be able to figure out what’s wrong and pin it down to individual components.
The authors argue the schizophrenia is too complex and varied to be examined in any way other than by breaking it down into smaller functional units – these are the ‘intermediate phenotypes’ that are also described as ‘endophenotypes’ by others. They list a number of these such as electrophysiology, neurochemistry and neuropsychology but really focus the review on what functional imaging studies have told us. Two brain circuits are described:

1) Cortex/thalamus > neostriatum > globus pallidus > substantia nigra > prefrontal cortex.

2) Dorso-lateral prefrontal cortex <> hippocampal formation

which are observable in action during these functional brain imaging studies. These circuits can be tied in with particular neuropsychological correlates (episodic memory, working memory, emotional regulation, reward) which in turn can be placed in the context of the features of schizophrenia (the first two) or depression (the last two). The authors explain how traditionally gene polymorphisms can be tested for their effect on these phenotypic systems allowing a prediction to be made whether they might have relevance to the disease state. With respect to their main hypothesis, the authors also show how gene polymorphisms can examined in an analogous way using their effect on the function of the two neural circuits as the intermediate phenotypic readout.

This is cutting edge technical stuff and the correlation between particular gene polymorphisms (e.g. in the DISC1 and COMT genes) and precise functional changes in particular brain regions has been nothing less than spectacular. Actually, this success is why I have certain conceptual issues with this approach. The kind of sample numbers (cases studied) involved are relatively low compared to the numbers employed in straight genetic studies trying to match up particular genetic changes with the disease state and yet the former are yielding statistically robust findings while the latter are most definitely not. This could be telling us one of three things:

1) Like the deleterious effects that many, many gene knockout experiments have on mouse learning and memory tests/electrophysiology, the biology of these circuits might be very susceptible to genetic changes.
2) The distance between gene polymorphism and intermediate phenotype is a lot less than between intermediate phenotype (or gene polymorphism) and disease. i.e. other genes/intermediate phenotypes are required in parallel before the sub-clinical becomes clinical.
3) Testing a pre-selected genotype for defined functional neural changes is a much purer experiment than asking whether your particular population of schizophrenia cases is enriched for that genotype (i.e. it avoids locus and allelic heterogeneity issues).

So there are several questions to be answered (as the authors acknowledge) such as the heritability of these functional brain activities, their extent of variation in the disease state, the contribution of other, less tractable neural circuits to disease and, most importantly, the effect of the emerging better-characterised pathogenic polymorphisms/haplotypes. Overall, however, this is a thought-provoking review which bodes well for a deeper understanding of the building blocks of mental illness – a subject discussed in a previous post.

Neurobiology of Schizophrenia
Ross, Margolis, Reading, Pletnikov and Coyle

Out of the three, this is perhaps the best review for the novice to dive into. It is an impressively comprehensive ‘State of the Nation’ piece for psychiatric disorders - as befits its publication in the prestigious journal Neuron: normally a place where schizophrenia and bipolar disorder don’t get a look in. Like its cousin, Cell, Neuron is a journal that, in my opinion, sometimes suffers from its preference for rather turgid, hard-core cell biology/electrophysiology. Does this mean that this field of psychiatric genetics is gathering a veneer of respectability?
The paper begins with a suggestion that we should look to the established genetics of neurodegenerative diseases such as Parkinson’s and Alzheimer’s to get a model for the genetic architecture of schizophrenia and bipolar disorder: rare mutations can tell us a lot about the biology of the disorders even if they don’t affect the majority of sufferers. They also suggest disorders of the brain such as Lissencephaly can be a model for the Neurodevelopmental Hypothesis of psychiatric illnesses.
This review also goes into the basics of endophenotypes, as described above, and also mentions the work into neuropsychology, neuroimaging, neuropathology, pathophysiology and recent progress in pharmacology. After introducing the readers to these fields, the review then shifts gear into a summary of genetic progress. This covers animal models and the key methodologies employed in the field for the identification of candidate genes: linkage, association and cytogenetics. 19 candidate genes are listed in their ‘Billboard’ chart. In fact, I’m embarrassed to say I didn’t even recognise a couple of the acts on this chart so there’s some homework for me….but, equally, I couldn’t see a few expected names either. I think the Schizophrenia Research Forum has a current list of ‘top’ candidate genes…..I wonder if it would be profitable for this blog to do so as well? Neuregulin, Dysbindin, D Amino Acid Oxidase Activator, COMT and DISC1 are described in more detail, particularly the advanced state of cell biological investigation into the last of these.

Overall, this is the current gold-standard review of the field – required reading for all. Just as important is the feeling that the reader gets when going through it: this is a field that is maturing nicely and with a lot to be positive about.

Here are a couple of links spotted on Metafilter that raise the issues surrounding how, and by whom, science papers are judged suitable for publication in journals. The first link discusses possible changes and the second highlights the strategy taken by the journal Nature.

Perhaps rather an esoteric subject, but one that occupies the minds of researchers a great deal.
Would a new reviewing process - perhaps based on ‘live’ editing/reviewing - help fix some of the problems associated with anonymous peer review?

1. It may help a small field like psychiatric genetics where papers are likely to be reviewed by a competitor. Live reviewing might dilute, or require justification of, criticism based on axe grinding.
2. Who is going to spend time on live review of small papers in the lower journals? I can’t see sparkling, erudite discussions on a the analysis of an over-employed polymorphism in a hackneyed candidate gene and its proposed effect on a minor pseudo-psychiatric phenotype.
3. When will a live reviewed/edited paper ever be deemed complete? Will it be ‘re-opened’ for butchery should opinions change over time?
4. Just like big editorial personalities can determine the content of a journal, will vociferous hawks push publications (and research) in unhelpful directions?

You may have seen the various press snippets (link 1, 2, 3, 4) surrounding the advanced online publication of a paper in the journal Nature this week (also comment in John Hawks’ Blog). This work is part of a field where bioinformaticians (those who study the large sets of data generated by projects such as the Human Genome Project and the HapMap Project) study the evolutionary footprints that are to be found in the DNA sequence of our chromosomes.

Since the Human Genome Project was completed five years ago, it has been joined by genome projects from a wide variety of inhabitants of the evolutionary tree so that it is possible to dissect the genetic differences between them. In the most part, this has been used to identify similarities between species. These ‘conserved’ regions have arisen because they have been subject to strong selective pressure not to accrue random mutations. Conserved regions usually indicate the presence of genes (which have to be kept intact in order to make functional protein) or regulatory regions (which control the use of particular genes). You can think of these conserved regions as being responsible for the core properties of a protein, and on a larger scale, responsible for the general features of an organism - head, limbs, organs, metabolism, muscle etc. It also stands to reason that mutations in these conserved regions are likely to have the most serious consequences……that of causing genetic disorders.

However, some researchers have looked at the flipside of this coin - hunting out the regions of the genome which show the greatest differences between species. These, it is hypothesized, represent the sites where natural selection and speciation events have occurred - they are responsible for what makes us different from our evolutionary ancestors.

This Nature paper details the application of this latter approach: the assumption that genes showing fast evolution between the homininae lineage (human, chimpanzee, gorilla and orang-utan) and hominina lineage (humans and their ancestors since the split with chimpanzees circa 6 million years ago) are responsible for our defining and distinguishing evolutionary features, in particular, increased brain size/cognitive power.

So, in brief, the reasoning goes like this: fast-evolved human genes > used in brain > they made us what we are today.

It’s not the first time that such an approach has been tried. Indeed, genes such as ASPM and Microcephalin have shown fast evolution and appear to play a role in brain development (lots of nice reviews here: link 1, link 2, link 3, link 4). There are even genes which are entirely specific to humans - for example the SIGLEC11 gene is expressed in brain microglial cells.

The first author of the paper, Katherine Pollard, carried out a systematic screen of the human genome, comparing it with chimp, in order to identify what she calls ‘Human Accelerated Regions’ (HARs) - the fast-evolved regions. The most significant finding was called HAR1 and shows 66 times more sequence changes than would be expected of an average stretch of DNA. The HAR1 region seems to be correlated with two overlapping genes, HAR1F and HAR1R. The genes themselves are odd in that they don’t make proteins - their function must be carried out in their mRNA form. Most RNA genes are regulatory in that they control the actions of other genes, but no such role has yet been shown for HAR1F/R. Again, the authors show that this humanised gene acts in the brain, and more specifically, the developing foetal brain. It is found in the precursor cells that will go on to produce the cerebral cortex - the higher functioning, business end of the brain.

How does schizophrenia fit into this story? A related hypothesis follows directly on from the ‘fast-evolved human gene’ argument. This states that schizophrenia is a uniquely human disease caused by dysfunction of uniquely human cognitive faculties and is, therefore, likely to have genetic causation in those brain genes that have driven human speciation. In other words, genes like the HARs are possible candidates for schizophrenia genes. Here are three papers that originally stated this idea.

Brune, M. (2004) Schizophrenia-an evolutionary enigma? Neurosci Biobehav Rev 28 (1), 41-53

Randall, P.L. (1998) Schizophrenia as a consequence of brain evolution. Schizophr Res 30 (2), 143-148

Crow, T.J. (1995) A theory of the evolutionary origins of psychosis. Eur Neuropsychopharmacol 5 Suppl, 59-63

It has to be said that this idea has been in the wilderness for a while, overtaken by more pharmacological thoughts on the genes involved in schizophrenia. Does it deserve another shot at glory? It’s hard to say just yet, but there are a few intriguing clues:-

1. HAR1F/R is found in the same cortex precursor cells as Reelin, itself a reasonably well-established schizophrenia candidate gene.
2. Another brain capacity gene, like ASPM and Microcephalin, is Nde1 (’Nude 1‘) which is a direct interactor with the DISC1 schizophrenia candidate gene.
3. A previous paper on fast-evolving brain genes identified GRIK4 (a glutamate responsive neurotransmitter receptor in the brain). We have just described its involvement in the genetics of schizophrenia and bipolar disorder.
4. Similarly, the NPAS3 gene was identified as ‘HAR21′ (also incorrectly labeled HAR30 in table S8) by Pollard et al. We have shown that this gene is involved in schizophrenia too - and described a region of the gene/protein that is human-specific (although this is not the same region flagged up in HAR21).

Circumstantial at best maybe, but certainly the basis for future study.

What an chillingly emotive idea, though. The genetic forces that boosted our cognitive faculties in Africa’s Rift Valley may contribute to the debilitating characteristics of psychiatric illness.

I wonder if this relatively unsung story could be the biggest advance in the genetics of psychiatric illness so far this year. It’s been a slow story building - I can remember hearing a talk given by Nicholas Barden, the French-Canadian head of the group behind this finding, at a conference in Dublin almost two years ago where he was on the verge of announcing the identity of the gene subject to its patent protection. Sometimes, the perceived importance of a gene can make the road to publication more rocky than usual. Certainly, it’s baffling to hear the hoops that some papers have to jump through to reach dusty library shelves. No doubt the jostling for position in this young field plays a part.

The story starts with the study of mental illness in isolated populations in the North-East of Canada. This is a genetic trick that has been employed widely in the study of modern complex human genetics. The ‘trick’ is that these populations usually start off with a few pioneers and if one carried a gene mutation then it will go on to be relatively common in the present population and will contribute to a larger proportion of the genetic risk. This approach has been used very effectively in Finland and Iceland.

Barden’s group identified a region on chromosome 12 which contained a gene mutation which increased the risk of bipolar disorder. The laborious task of refining the location of the gene led to the pinpointing of P2RX7 (a.k.a. P2X7) as the most likely source of the risk. Indeed, a change in DNA sequence (with the reference number rs2230912)which leads to a change in the amino acid sequence (a glutamine to an arginine residue) of the P2RX7 protein was found.

This is what sets this finding apart: many candidate genes have been found that seem to increase risk of mental illness but the underlying mutations have not yet been found (e.g. NRG1). Add to that a number of gene mutations which seem to remain as genetic candidates more through tearful sentimentality and half-hearted dabbling rather than thorough replication (COMT), then it can be seen what it means to have both gene and mutation under your belt.

The statistical power of the studies is another favourable sign. Their first paper, from earlier this year, looking at families with bipolar disorder in the French-Canadian population found that the mutation was significantly associated with the disease. This has very recently been backed up by a study with German colleagues which has found a similarly pronounced effect of the mutation on disease risk through the study of 1000 individuals with unipolar depression.

So much for the genetic story, but what exactly does the P2RX7 protein do, and what relevance does it have for the biology (or treatment) of bipolar disorder and unipolar depression? Here the story becomes a little less clear. Adenosine TriPhosphate (ATP) is the cell’s universal energy currency, converted from sugar energy by mitochondria. Many enzymatic processes (e.g. muscle contraction) in the cell rely on ATP power. However, recently it has been appreciated that ATP can also act as a signalling molecule outside of the cell. A number of classes of genes encode proteins which act as ATP receptors (detectors) on the cell surface. The P2RX class of receptors is one of these and consists of seven members, including P2RX7, which let ions pass across the cell membrane when they sense the presence of ATP. Online Mendelian Inheritance in Man (OMIM, a good resource for summaries on gene function) has already got a substantial entry for P2RX7, one that focuses more on its role in leukaemia, inflammation and diabetes. This is mirrored in a more up-to-date and comprehensive review which discusses P2RX7’s role in arthritis and other inflammatory disorders. As is so often the case for a given gene, the field that discovers it describes it only in its own terms. Thus mice with the gene ‘knocked out’ have been studied in terms of their resistance to arthritis and bone growth deficits. However, there is an emerging literature (1, 2, 3, 4) which suggests a role for this receptor in the brain and, potentially, with glutamate release in the hippocampus.

A final point of interest is that the harmful effect of this mutation is only felt when in the heterozygous state (one good copy of the gene and one bad) which results in the production of 50% normal and 50% abnormal protein. The fact that P2RX7 protein acts in a homomeric complex (it binds to other P2RX7 proteins to form the active receptor), and the changed amino acid resides in the part of the protein responsible for this self-association suggests that the normal-normal and abnormal-abnormal associations are non-harmful but the abnormal-normal combination is.

Peer-reviewed papers in respectable journals are to press releases what spring water is to cheap sparkling wine. Trawling the interweb, I came across this. I’m afraid you have to register to view the full article but it basically distills down to an early fanfare for the upcoming production of a home test for schizophrenia. The article describes very little about the product itself but mentions the company’s current market research drive to work out how to sell it.

“We’re not promising a cure…….What we hope to do is provide families with an accurate picture of their risk.”

The company itself, SureGene, has a rather dowdy web-site which mentions patented LocusLock(TM) technology which appears to be a linkage method which outputs locus interaction information not normally generated by such studies.
Upon further investigation, I find once more that I am several months behind the rest of the World and that the implications of this test have already been discussed and commented on by others: include some very courteous and moderate responses from a SureGene manager. The company appears to have published one schizophrenia gene candidate, SULT4A1, which seems to have been found through other means but its significance is unclear at present.

Three questions spring to mind:-

1. What is the nature of the ‘kit’ itself. I suspect it is a tube for the colection of saliva which is then shipped back to the company for DNA typing. I can’t imagine what protein target or bodily fluid might currently be suitable for an alternative ‘dip-stick’ type of schizophrenia diagnostic kit (e.g. hCG and urine for pregnancy tests).
2. Because schizophrenia is genetically complex, shows incomplete penetrance, and differs in cause between unrelated individuals, what might a DNA test be looking for? They seem to imply that their LocusLock(TM) technology has identified a network of interacting candidates….a better approach than looking at a single gene but still hit-or-miss. For instance, they may know the contributory genes in an ideal world, but do they know (and have assays for) the underlying mutations?
3. What are the ethical and personal issues surrounding such a test? Would the test result be presented as a definitive finding or reveal its status as a calculated assessment of % risk? What life-changing decisions would be based on a test result? How many Canadian ‘online meds’ sites would spring up, a la Tamiflu, to supply a booming market for preventative anti-psychotic drug regimes? BuUyyy C#e@p 0l@nz@p1ne spam, anyone?

Time and the market-place will determine this and other attempts at diagnostics. Personally, I think the science isn’t there yet….but I think this should serve as a wake-up call for the genetic counsellors, at least.

A Quantitative Trait (QT) is a phenotype/feature/measurement of an organism that can fall anywhere on a continuous range. So height, intelligence, metabolic rate, longevity etc. all sit within this class of traits. On the other hand, there are traits such as eye colour, blood group, tongue rolling etc. which are qualitative and discrete in nature. As common sense might dictate, the distinction between these two sets of traits is mirrored in the complexity of the genetics behind them. Many genes are predicted to simultaneously contribute to quantitative traits and these genes are termed QT Loci (QTLs: loci referring to the locations of these genes on the chromosomes). A handful of genes at most are thought necessary to control a qualitative trait.
A very recent paper from Jonathan Flint’s laboratory (Oxford, UK) describes a large mouse genetics study that aimed to uncover the genetic contributions to a number of measurable traits (including behavioural, metabolic, weight, biochemical, blood properties etc.).

The considerable advance described in this paper was the scale of the project coupled with the use of what they call heterogeneous stock mice. Basically these are derived from the interbreeding of a different original mouse strains (each with their own particular trait characteristics) so that, after many generations, mice are produced with an apparently random mix of these traits caused by the shuffling and mixing of the original genetic material on their chromosomes. By using markers which can tell which parts of which chromosomes came from which original mouse strain it was possible, with necessarily powerful and complicated statisitical methods (I don’t understand them so won’t attempt to summarise them!), to work out the approximate location and contribution of many QTLs for a given trait. The exact identity of the QTLs will doubtless follow in subsequent publications.
The upshot of this gargantuan effort is that, for the average trait, many QTLs are responsible, but each one is unlikely to contribute to much more than 5% of the trait and that only 75% of the genetic effect on the trait could be pinned down to detectable QTLs (i.e. many other undetectable QTLs exist, each contributing very small amounts to the final trait measurement).

What are the implications of this work? The authors have achieved the first steps in making the identification of QTLs a realistic possibility. However, the link to human disease is a little harder to make.

Of course, I have a vested interest in hoping that the QTL story is not relevant to the identification of genes involved in mental illness because it would make it a much harder task. And in that respect, there are also a few conceptual issues I have with equating QTLs with diasease genes.

Firstly, these mice are all well. A few years back I heard a suggestion that the genes for intelligence could be cloned by identifying the genes responsible for mental retardation (UK: learning disability). Yes, both intelligence and mental retardation can be defined in terms of IQ but the comparison prety much ends there. An analogy illustrates the problem: if the speed of a car represents IQ then it is possible to list the modifications/properties (QTLs) which might alter the performance - engine capacity, tyre type, turbo-charging, air intakes, computer tuning etc. etc. However, contrast this with the things that can go wrong - fan belt broken, accelerator pedal snapped, puncture, oil leak etc. etc. - and bear in mind that these factors play only minor roles in determining the car’s ‘performance’ per se. Despite this being only an analogy, we should not be surprised if natural variations in the ’state of wellness’ have little bearing on common illness.
Secondly, the genes responsible for mental illness (or any other complex genetic disorder, for that matter) exist as a consequence of spontaneous mutations in genes which have then been distributed and selected among populations which have then been scattered by numerous ancient migrations. All these factors have determined the current set of mental illness gene variants present at reasonable frequencies in our human populations. This is also true for the specific QTLs present in the parental strains of mice used in these experiments. Whether there is likely to be susbtantial overlap in contributory genes for any trait/disease between mouse and man should be a subject for debate rather than an assumption.

In summary, is mental illnes (in an individual rather than a population) caused by small-effect changes in many genes acting in concert or by the actions of one or few genes with stronger effects? I believe the current state of play, as evidenced by the emerging genes, suggests the latter is more likely. Indeed, the current techniques such as linkage and case-control association studies employed in the search for these genes have a definite lower limit in their sensitivity which means that we have to hope that this is the case. I guess the fact that those are the kinds of genes we are seeing published recently may just be a result of the use of these techniques - ascertainment bias as it is known - but there is nothing to suggests that we are looking solely at a QTL effect.
Next post I will discuss another recent paper which describes further work linking a variant of a gene and its link with bipolar disorder and major depression: a potentially important major candidate for these disorders.

Here are some interesting podcasts/transcripts from episodes of the US radio broadcast series ‘All in the mind‘.

1. Valium
2. Schizophrenia 2
   
3. Mental illness and art
4. Depression
5. Hearing voices
6. Schizophrenia 1
7. Omega-3 fatty acids

…and plenty more to peruse.

The comedian Stephen Fry is to present an account of his life with bipolar disorder….interesting to read that he had never heard of the condition before his diagnosis. At the risk of appearing ghoulish or sensationalist, here is a list of well-known idividuals with bipolar disorder. If nothing else, this kind of thing serves to remind us of the indiscriminate nature of psychiatric illness.

“A little learning is a dangerous thing; drink deep, or taste not the Pierian spring: there shallow draughts intoxicate the brain, and drinking largely sobers us again.“

Alexander Pope - An Essay on Criticism, 1709

It is often the young fields like psychiatric genetics which suffer from over-interpretation in the face of minimal facts. I previously worked in the field of DNA methylation and the often-whispered joke was that there were more published reviews on the subject than primary research papers. Psychiatric Genetics is definitely showing a similar trend too.

I have no probems with this state of affairs in the most part. Everyone needs a digestable point of entry into a new field and a covenient method to delve deeper into particular aspects. However, there is an emerging trend in certain papers (reviews and others) to construct biological models of schizophrenia, bipolar disorder or other complex disorder biology based upon aggregated data. Inevitably, these models are used to add weight to candidate gene findings or to suggest novel candidates.

Over-cooking the Bacon

You may be familar with the thought game ‘six degrees of Kevin Bacon‘ or its intellectual predecessors. Basically, the theory behind this can be summarised as ‘because things are connected (in many different ways) to many other things, it doesn’t take many steps to connect any one thing to any other thing’. ‘Things’, in this context, can refer to citizens of the World, film actors or - perhaps -genes in the genome.

In the case of genes, the ways they can be connected with other genes is manifold:

1. they look the same (sequence homology)
2. they participate in a similar process
3. they are found in the same tissue or sub-cellular region
4. they are regulated by the same stimuli
5. they are expressed at a similar developmental time-point
6. they directly interact (protein binding, enzyme/substrate interaction)

A number of bioinformatics approaches have attempted to construct networks of these interactions. These have been based on trawling through the databases of published scientific papers in order to extract ‘co-incidences’ or on the grouping together of genes according to keyword definitions (gene ontologies) assigned to each gene. These bioinformatic strategies have principally been used to suggest genes involved in a given disease/process and to provide meaningful interpretation to data-spawning experiments such as expression microarrays (a test which looks at how a state or stimulus alters the use of a large set of genes).

So there have been publications with titles like ‘Post-mortem studies on schizophrenic brain gene expression imply changes in genes involved in X, Y and Z processes’ or ‘Bioinformatic approaches demonstrate that schizophrenia is a disorder of p and q pathways’. And this is where I have conceptual difficulties in these approaches - not necessarily with the veracity of the results - but in trying to understand how they can contribute to the future development of the psychiatric genetics.

Green screen - recycling old knowledge

Two problems spring to mind with this methodology. Firstly, the ’six degrees of molecular interaction’ issue introduced above suggests that ‘over-linking’ of genes into apparently signficant pathways is probable. This ‘noise’ would cloud the specific ’signal’ relevant to the study’s aim. Secondly, there is no additional information generated in the bioinformatic cataloguing beyond the original published data - and no easy means to ensure that the source data is of uniform quality and format. Therefore, although some quantitative evidence will result from the interrogation of the data, no new qualitative findings can result - you basically get out what you put in and nothing more. If gene S (causing schizophrenia) isn’t on your microarray chip, wasn’t in your yeast two-hybrid expression library or doesn’t interact with your chosen bait, is expressed at low levels in the brain, shares no homology with other genes or, heaven forbid, doesn’t fit into previously hypothesised biological pathways implicated in schizophrenia….then you are not going to find it, however much you shake up the input data.
If any readers have examples where this approach has yielded unexpected, novel findings which have been confirmed by specific experimental research then I would be keen to hear about it…and eat my words!

Having said all that…

Having said all that, if I had the opportunity and means to carry out this kind of work or to use its consequences to justify my own findings, then I would do so in a shot. Clutching at impressive straws is the psychiatric genetics way.