Jinmyung Choi , Parisa Shooshtari , Kaitlin E Samocha , Mark J Daly , Chris Cotsapas
doi: http://dx.doi.org/10.1101/017277
AbstractInfo/HistoryMetrics Preview PDF
Abstract
Using robust, integrated analysis of multiple genomic datasets, we show that genes depleted for non-synonymous de novo mutations form a subnetwork of 72 members under strong selective constraint. We further show this subnetwork is preferentially expressed in the early development of the human hippocampus and is enriched for genes mutated in neurological, but not other, Mendelian disorders. We thus conclude that carefully orchestrated developmental processes are under strong constraint in early brain development, and perturbations caused by mutation have adverse outcomes subject to strong purifying selection. Our findings demonstrate that selective forces can act on groups of genes involved in the same process, supporting the notion that adaptation can act coordinately on multiple genes. Our approach provides a statistically robust, interpretable way to identify the tissues and developmental times where groups of disease genes are active. Our findings highlight the importance of considering the interactions between genes when analyzing genome-wide sequence data.
Haldane's Sieve 
Doesn’t seem like a particularly new insight to me. If you take all known haploinsufficient disease genes and just look at their associated clinical phenotype you’ll see a major enrichment for neurodevelopmental disorders. Would be very odd if constrained genes were not enriched in fetal brain development.
The idea is nice, but not novel. This seems like a more thorough version of the approach:
http://genome.cshlp.org/content/25/1/142.short
However, I do like the implementation in this work. It would be nice if something like this could be done for common variants? I guess we don’t have as strong of assumptions about how they ought to group together though.
@Matt – just to be clear, we make no claims about haploinsufficiency, but about selective constraint. The novelty, as I see it, is not that constrained genes are involved in early fetal brain development – that’s expected (though expected and thoroughly proven aren’t quite the same thing). Rather, it’s that constraint acts on sets of interacting genes (the constrained genes aren’t a random subset of early fetal brain genes, for instance). I like to think the robustness of the approach might be useful and interesting too.
And yes, of course it makes sense 🙂
@James – we have been thinking about common variants quite a lot (that’s our bread and butter) and in fact had originally developed this approach for that application. Expect to hear more in the near future – if we ever deal with LD to my satisfaction.
Pingback: Most viewed on Haldane’s Sieve: April 2014 | Haldane's Sieve