in protein evolution, and therefore macrobiotic speciation, are now well understood, and closely match the timescales documented in the fossil record 5 . The elephant in the room for many genetic-algorithm opti- mization processes is therefore that the number of generations required for optimization may be high, whereas the possible throughput of experiments is often low. Kreutz et al. 1 have combined solu- tions to all of the above issues into a system that holds real promise for developing useful answers to syn- thetic problems involving multi- variate optimizations. The authors studied the oxidation of methane (CH 4 ) to methanol (CH 3 OH, an industrially important bulk chemi- cal) using oxygen gas. This process is of increasing interest because of its environmentally friendly creden- tials; oxygen is readily available from the atmosphere, and produces no unpleasant waste products, unlike many other oxidants. But catalyst turnover numbers for the reaction (the number of moles of substrate converted by each mole of the catalyst before it is inactivated) have been low, as have the yields in many cases. The authors therefore designed a droplet- based microfluidic system that used a genetic algorithm to optimize the properties of a meth- ane-oxidation catalyst. The catalyst consisted of three components, each of which could be varied. The team began by screening 48 differ- ent combinations of components for their col- lective catalytic activity, with every combination tested in its own droplet. Each combination consisted of three ‘genes’, sets of compounds that might act as one of the three components of the catalyst. The authors identified the best genes (those that produced the most active catalysts) and either ‘mutated’ them by chang- ing a few of their components at random, or crossed them with other genes. This gen- erated a new set of 48 genes for a subsequent round of screening. Over eight generations of screening, the team generated an up to sevenfold increase in the fitness score — a combination of yield and turnover number — of their catalytic sys- tem, and also identified the relevant catalytic components that enhanced fitness. This is an amazing achievement given the relatively small number of component combinations (384) explored. For comparison, there are hundreds of millions of possible combinations. The method is also notable because of the consummate use of the properties of both the techniques and the materials. The chip mater- ial chosen by Kreutz et al. was polytetrafluoro- ethylene (Teflon), a fluorous material in which all the hydrogens of the hydrocarbon chains have been replaced with fluorine atoms. Similarly, the droplets were carried through the microfluidic microfluidic device. Kreutz et al. have thus directly addressed — and, in the case of droplet crosstalk, har- nessed — the three elephants in the room associated with microfluidic devices and genetic algorithms. This combination of techniques and materials 1 opens up the possibil- ity of the evolutionary development of other, more complex catalysts and even of artificial enzymes. The appli- cations are therefore widespread within biomolecular and chemical sciences. However, challenges do remain. Notably, although gases readily permeate Teflon, liquids do not. Similarly, the rate of mass trans- fer through the walls of the droplet- carrying tubes is limited by the wall’s permeability, thickness and the pressure differential across it. These effects might sometimes combine to cause a paucity of starting materials, artificially slowing fast reactions. Regardless, it will be fascinating to see what discoveries will be made as a result of Kreutz and colleagues’ approach. Rarely has a herd of elephants seemed so productive. Robert C. R. Wootton is at the School of Pharmacy and Biomolecular Sciences, Liverpool John Moores University, Liverpool L3 3AF, UK. Andrew J. deMello is in the Department of Chemistry, Imperial College London, London SW7 2AZ, UK. e-mails: r.c.wootton@ljmu.ac.uk; a.demello@imperial.ac.uk 1. Kreutz, J. E. et al. J. Am. Chem. Soc. 132, 3128–3132 (2010). 2. Lee, J., Kim, M. J. & Lee, H. H. Langmuir 22, 2090–2095 (2006). 3. Badal, M. Y., Wong, M., Chiem, N., Salimi-Moosavi, H. & Harrison, D. J. J. Chromatogr. A 947, 277–286 (2002). 4. Huebner, A. et al. Lab Chip 8, 1244–1254 (2008). 5. Hedges, S. B. & Kumar, S. Trends Genet. 19, 200–206 (2003). device by a fluorous oil. Such fluorous materials are extremely permeable to gases, particularly oxygen, so the authors simply perfused the gaseous starting materials for the reactions through the walls of the chip (Fig. 1). Kreutz and colleagues monitored the pro- duction of methanol in their system using droplet crosstalk. In this approach, catalyst- bearing droplets were flanked by droplets containing colorimetric indicators. These indi- cators responded to methanol diffusing from adjacent catalyst-bearing droplets by changing colour, thus allowing a rapid, on-line assess- ment of reaction progress. What’s more, the authors proved that their genetic algorithm was practically useful in a high-throughput Figure 1 | Screening catalysts in a microfluidic device. Kreutz et al. 1 have designed a microfluidic system that screens the activity of catalysts for the oxidation of methane (CH 4 ) to methanol (CH 3 OH), using oxygen as the oxidant. Microdroplets of catalyst solutions are passed through Teflon tubes by a carrier fluid. The Teflon tubes are enclosed by stainless-steel tubes, into which pressurized CH 4 and O 2 are introduced. The gas mixture diffuses through the Teflon tubing and, in the presence of active catalysts, reacts to produce methanol (CH 3 OH). At high temperatures, the methanol diffuses through the carrier fluid to a neighbouring droplet, which contains a colorimetric indicator. The indicator changes from yellow to purple in the presence of methanol, with the intensity of the purple colour corresponding to the activity of the catalyst. (Figure adapted from ref. 1.) EVOLUTIONARY BIOLOGY A flourishing of fish forms Michael Alfaro and Francesco Santini According to an innovative exercise in ‘morphospace analysis’, modern fish owe their stunning diversity in part to an ecological cleaning of the slate by the mass extinction at the end of the Cretaceous. Fish come in a bewildering diversity of shapes: trumpetfish, sea horses, pufferfish, cowfish and anglerfish are just a few exam- ples. Size, too, varies tremendously in living species. Some gobies reach less than a centi- metre as adults, whereas the oarfish stretches well over 12 metres and the ocean sunfish can weigh more than a car. These groups, along with almost two-thirds of all other fish, belong to a particularly conspicuous section of the fish tree of life, the spiny-finned fishes (Acanthomorpha). With 18,000 species, spiny- fins comprise almost a third of all vertebrates. Despite, or perhaps because of, this incredible richness, biologists understand surprisingly little about the tempo of spiny-fin diver- sification. As he reports in Proceedings of the Royal Society, Friedman 1 has searched for the origins and underlying causes of shape diversity in spiny-fins across nearly 840 NATURE|Vol 464|8 April 2010 NEWS & VIEWS © 20 Macmillan Publishers Limited. All rights reserved 10