One of last week’s meetings (that we co-sponsored) was an all-day discussion at the RAC Club on the future of agriculture and food production, organised and skilfully facilitated by Sir Ben Gill and entitled From a Land of Plenty to a Land of Uncertainty. Sub-themes were (i) the competing demands for land use, (ii) waste, and (iii) water. I mentioned waste a little last week, and so I shall devote some space here to water, a commodity that it is widely considered will be insufficiently available as the need for agricultural productivity increases through demographics exacerbated by climate change.

Agriculture uses huge amounts of water (one acre-inch is 22,000 gallons), but most of the water we use in food is virtual, hidden or ‘embedded’, with for example the growth of the beans needed to make a cup of coffee requiring some 148 litres, and 300 tonnes of water needed to make the 8 tonnes of wheat that is the approximate UK yield from 1 hectare. Most of UK agriculture relies on rainfall rather than controlled irrigation, but some areas (e.g. California) do irrigate heavily.

John Gummer spoke on a variety of needs that he had identified, including the reform of Ofwat, water metering coupled to social circumstances, the need for water storage to be improved considerably, the need to seek local solutions, and a recognition that coastal erosion does not properly value agricultural land. There are many market failures, some self-inflicted in that while it would be cheaper for water companies to pay farmers not to overuse nitrates, cross-subsidy is not allowed. Equally odd is the fact that presently the energy value of wheat is said to be £140/tonne, while the food value (market price) is £100/t. As we surely discovered from the credit crunch, the ‘efficient market hypothesis’ is found wanting.

When we consider electricity distribution, it is well understood that a ‘smart grid’, where detailed monitoring and control systems adjust both supply and demand intelligently, will be part of the solution. I see no reason why a smart water grid or network should not be an important part of the solution for water distribution.

Another kind of network that is of huge importance is of course the internet, which for UK academics and research councils means Janet. This is paid for via JISC, and I attended my first and very interesting meeting of the JISC Board. Working out our needs for bandwidth is going to be an important part of ensuring that they do not limit our ambitions as we move to an age of Big Data. It is well worth reflecting on the fact that the UK’s success in attracting the European Bioinformatics Institute to our shores was due to our vastly superior academic network bandwidth relative to that of the competition in continental Europe.

One of my scientific interests is in the effective use of combinatorial methods to explore complex and high-dimensional spaces. If only 1% of mutations increase fitness then only 1 in 10,000 double mutations will (if independent), so most mutational schemes in directed evolution experiments use a rather low mutation rate equivalent to 1 per object. Actually, there is a flaw in this logic, since what matters is the magnitude of the improvement when it occurs, and it turns out (as reviewed) that much higher mutation rates are often more effective in increasing fitness. A recent in vivo experiment with phage T7 reinforces this conclusion, where increasing natural mutations rates by 2-3 logs over baseline increased fitness in an adaptive manner. The optimal mutation rates depend on many factors, but there is usually at least one such rate.

A related question, more at the level of network biology, pertains to finding a small number of specific genetic changes that together might offer a huge improvement in some trait of interest. If there are 3 such changes in 1000 proteins, the number of combinations (of 3 out of 1000) is ca 100 billion, so experimental methods will struggle to find them. By contrast, in silico methods can find them easily, as in the improvement in valine productivity described by Sang Yup Lee and his colleagues. Equally, when one has the variation it is then possible to pin down the relevant genes, and a nice example of this was provided in a recent paper that showed that almost all the variations that together account for most coat phenotypes in purebred dogs in the USA were controlled by just 3 genes. As the authors comment, “an array of varied and seemingly complex phenotypes can be reduced to the combinatorial effects of only a few genes”. The task of genomics-driven breeding is to find and exploit them.

Related posts (based on tags and chronology):