Showing papers by "Suet Yi Leung published in 2010"

Prognostic relevance of DNA copy number changes in colorectal cancer

Abstract: In a study of 109 colorectal cancers, DNA copy number aberrations were identified by comparative genomic hybridization using a DNA microarray covering the entire genome at an average interval of less than 1 Mbase. Four patterns were revealed by unsupervised clustering analysis, one of them associated with significantly better prognosis than the others. This group contained tumours with short, dispersed, and relatively few regions of copy number gain or loss. The good prognosis of this group was not attributable to the presence of tumours showing microsatellite instability (MSI-H). Supervised methods were employed to determine those genomic regions where copy number alterations correlate significantly with multiple indices of aggressive growth (lymphatic spread, recurrence, and early death). Multivariate analysis identified DNA copy number loss at 18q12.2, harbouring a single gene, BRUNOL4 that encodes the Bruno-like 4 splicing factor, as an independent prognostic indicator. The data show that the different patterns of DNA copy number alterations in primary tumours reveal prognostic information and can aid identification of novel prognosis-associated genes.

Risk Assessment of Residual DDTs in Freshwater and Marine Fish Cultivated Around the Pearl River Delta, China

Abstract: Six species of freshwater fish collected from 10 fishponds in Shunde and Zhongshan, China, four species of marine fishes collected from different mariculture farms [four in Hong Kong (Tung Lung Chau, Ma Wan, Cheung Chau and Kat O) and two in mainland China (Daya Bay and Shenzhen)] together with feed (both trash fish and commercial pellets) and sediment were analyzed for DDTs. Total DDTs in freshwater fish flesh decreased in the order of: carnivores [1742 μg/kg lipid weight (l.w.)] > herbivores (165 μg/kg, l.w.) > omnivores (42.5 μg/kg, l.w.), with the highest concentration detected in mandarin fish (Siniperca chuatsi) (2641 μg/kg, l.w.). For marine fish, snubnose pompano (Trachinotus blochii) and orange-spotted grouper (Epinephelus coioides) collected in Ma Wan contained elevated levels of total DDTs (2590 and 2034 μg/kg l.w., respectively). Trash fish used in both freshwater and marine fish farms contained significantly higher levels (86.5–641 μg/kg l.w.) (p < 0.05) of DDTs than in commercial pellets, but correlations between DDT levels in fish feed and muscle were not significant. The elevated biota-sediment accumulating factor for tilapia (Tilapia mossambicus) (24.1) indicated that accumulation of DDTs from sediment to the fish was evident. It can be concluded that trash fish should not be used for fish culture in order to lower the level of residual DDTs in fish muscle.

Partially connected feedforward neural networks on Apollonian networks

Abstract: This paper presents a novel and data-independent method to construct a type of partially connected feedforward neural network (FNN). The proposed networks, called Apollonian network-based partially connected FNNs (APFNNs), are constructed in terms of the structures of two-dimensional deterministic Apollonian networks. The APFNNs are then applied in various experiments to solve function approximation, forecasting and classification problems. Their results are compared with those generated by partially connected FNNs with random connectivity (RPFNNs), different learning algorithm-based traditional FNNs and other benchmark methods. The results demonstrate that the proposed APFNNs have a good capacity to fit complicated input and output relations, and provide better generalization performance than traditional FNNs and RPFNNs. The APFNNs also demonstrate faster training speed in each epoch than traditional FNNs.

Predicting metabolic pathways from metabolic networks with limited biological knowledge

Abstract: Understanding the metabolism of new species (e.g. endophytic fungi that produce fuel) have tremendous impact on human lives. Based on predicted proteins and existing reaction databases, one can construct the metabolic network for the species. Next is to identify critical metabolic pathways from the network. Existing computational techniques identify conserved pathways based on multiple networks of related species, but have the following drawbacks. Some do not rely on additional information, so only locate short (of length at most 5), but not necessarily interesting, conserved paths. The others require extensive information (the complete pathway on one species). In reality, researchers usually know only partial information of a metabolic pathway and may not have a conserved pathway in a related species. The Conserved Metabolic Pathway (CMP) problem is to find conserved pathways from the networks with partial information on the initial substrates and final products of the target pathways. Experimental results show that our algorithm CMPFinder can predict useful metabolic pathways with acceptable accuracy.