Rob J Hyndman

Dav­ide Carta, Laura Vil­lan­ova, Stefano Cost­a­curta, Aless­andro Patelli, Irene Poli, Simone Vezzu, Paolo Sco­pece, Fabio Lisi, Kate Smith-​​Miles, Rob J Hyndman, Anita J. Hill, and Paolo Falcaro

Ana­lyt­ical Chem­istry (2011), 83(16), 6373–6380.

ABSTRACT: In industry as well as many areas of sci­entific research, data col­lec­ted often con­tain a num­ber of responses of interest for a chosen set of explor­at­ory vari­ables. Optim­iz­a­tion of such mul­tivari­able mul­tire­sponse sys­tems is a chal­lenge well suited to genetic algorithms as global optim­iz­a­tion tools. One such example is the optim­iz­a­tion of coat­ing sur­faces with the required abso­lute and rel­at­ive sens­it­iv­ity for detect­ing ana­lytes using devices such as sensor arrays. High-​​throughput syn­thesis and screen­ing meth­ods can be used to accel­er­ate mater­i­als dis­cov­ery and optim­iz­a­tion; how­ever, an import­ant prac­tical con­sid­er­a­tion for suc­cess­ful optim­iz­a­tion of mater­i­als for arrays and other applic­a­tions is the abil­ity to gen­er­ate adequate inform­a­tion from a min­imum num­ber of exper­i­ments. Here we present a case study to eval­u­ate the effi­ciency of a novel evol­u­tion­ary model-​​based mul­tire­sponse approach (EMMA) that enables the optim­iz­a­tion of a coat­ing while min­im­iz­ing the num­ber of exper­i­ments. EMMA plans the exper­i­ments and sim­ul­tan­eously mod­els the mater­ial prop­er­ties. We illus­trate this novel pro­ced­ure for mater­i­als optim­iz­a­tion by test­ing the algorithm on a sol
gel syn­thetic route for pro­duc­tion and optim­iz­a­tion of a well stud­ied amino-​​methyl-​​silane coat­ing. The response vari­ables of the coat­ing have been optim­ized based on applic­a­tion cri­teria for micro– and macro-​​array sur­faces. Spot­ting per­form­ance has been mon­itored using a fluor­es­cent dye molecule for demon­stra­tion pur­poses and meas­ured using a laser scan­ner. Optim­iz­a­tion is achieved by explor­ing less than 2% of the pos­sible exper­i­ments, res­ult­ing in iden­ti­fic­a­tion of the most influ­en­tial com­pos­i­tional vari­ables. Use of EMMA to optim­ize con­trol factors of a product or pro­cess is illus­trated, and the pro­posed approach is shown to be a prom­ising tool for sim­ul­tan­eously optim­iz­ing and mod­el­ing mul­tivari­able mul­tire­sponse systems.

Online paper