RESUMO
RESUMO O Sistema Cantareira (SC), um dos maiores sistemas produtores de água do Brasil, enfrentou um período de forte seca entre os anos de 2013 e 2014, o que comprometeu não só o atendimento a 9 milhões de pessoas na Região Metropolitana de São Paulo (RMSP) como também as bacias a jusante. Este estudo tem o objetivo de propor modelos de programação dinâmica (PD) para operação ótima do SC, considerando o período crítico de 2011 a 2015. Nesse contexto, foram formulados um modelo de otimização livre e um modelo de otimização baseado em uma regra de operação por faixas. Ambos os modelos apresentaram desempenho superior à operação realizada quanto ao volume armazenado e quanto ao atendimento às demandas. O modelo de otimização livre apresentou fornecimento médio de 29,40 m3.s-1 para RMSP e de 9,30 m3.s-1 e de 18,70 m3.s-1 para as bacias a jusante, representadas pelos pontos de controle Jb e Jv, respectivamente. A regra de operação por faixas forneceu vazões médias de 28,50 m3.s-1 para a RMSP, de 9,40 m3.s-1 para Jb e de 19,30 m3.s-1 para Jv.
ABSTRACT The Cantareira System (CS), one of Brazil's biggest water producing systems, went through a period of severe drought from 2013 to 2014, compromising not only the 9 million people in the Metropolitan Region of São Paulo (MRSP), but also the downstream basins. This paper aims to propose Dynamic Programming (DP) models for optimal operation of the CS, considering the critical period from 2011 to 2015. In this context, an unconstrained operation optimization model and a rule-based operation optimization model were formulated. Both optimized models presented superior performance when compared to the real operation regarding final storage volumes and guaranteeing the attendance of water demand. The unconstrained operation optimization model provided average supply of 29,40 m3.s-1 for MRSP, and of 9,30 m3.s-1 and 18,70 m3.s-1 for the downstream basins represented by control points, Jb and Jv respectively. The rule-based operation model provided average supply of 28,50 m3.s-1 to MRSP, of 9,40 m3.s-1 to Jb, and of 19,30 m3.s-1 to Jv.
RESUMO
Objective To predict signal transduction pathways of the integrin adhesome via bioinformatics method, so as to provide references for experimental study on the mechanism of integrin-related signal transduction pathways. Methods First, the interaction network between the interactive integrin adhesome was constructed, and the confidence probability of each interaction was used as its link weight, respectively. Secondly, the pathways of the minimum weight were identified via a standard dynamic programming algorithm. Finally, all pathways calculated by the algorithm were aggregated into some probable networks. Results Seven signal transduction pathways were obtained from the integrin adhesome interaction network that contained 147 components with 736 interactions. In every predicted signal transduction pathway, the coverage rate of the proteins with Gene Ontology annotation was calculated. Conclusions By research on signal transduction pathways the pathogenesis of some diseases can be explored at the molecular level. Several possible signal transduction pathways obtained in this study have some reference value for exploring disease mechanism in basic medical sciences, and also provide some useful information for such exploration under the stimulation of external signals including mechanical or chemical signals.
RESUMO
Local alignment is an important task in molecular biology to see if two sequences contain regions that are similar. The most popular approach to local alignment is the use of dynamic programming due to Smith and Waterman, but the alignment reported by the Smith-Waterman algorithm has some undesirable properties. The recent approach to fix these problems is to use the notion of normalized scores for local alignments by Arslan, Egecioglu and Pevzner. In this paper we consider the problem of finding all local alignments whose normalized scores are above a given threshold, and present a fast heuristic algorithm. Our algorithm is 180-330 times faster than Arslan et al.''s for sequences of length about 120 kbp and about 40-50 times faster for sequences of length about 30 kbp.