Jorge Vergara

A review of feature selection methods based on mutual information

In this work, we present a review of the state of the art of information-theoretic feature selection methods. The concepts of feature relevance, redundance, and complementarity (synergy) are clearly defined, as well as Markov blanket. The problem of optimal feature selection is defined. A unifying theoretical framework is described, which can retrofit successful heuristic criteria, indicating the approximations made by each method. A number of open problems in the field are presented.

The expression of syndecan-1 and -2 is associated with Gleason score and epithelial-mesenchymal transition markers, E-cadherin and β-catenin, in prostate cancer

The epithelial-mesenchymal transition (EMT) is considered a key step in tumor progression, where the invasive cancer cells change from epithelial to mesenchymal phenotype. During this process, a decrease or loss in adhesion molecules expression and an increase in migration molecules expression are observed. The aim of this work was to determine the expression and cellular distribution of syndecan-1 and -2 (migration molecules) and E-cadherin and beta-catenin (adhesion molecules) in different stages of prostate cancer progression. A quantitative immunohistochemical study of these molecules was carried out in tissue samples from benign prostatic hyperplasia and prostate carcinoma, with low and high Gleason score, obtained from biopsies archives of the Clinic Hospital of the University of Chile and Dipreca Hospital. Polyclonal specific antibodies and amplification system of estreptavidin-biotin peroxidase and diaminobenzidine were used. Syndecan-1 was uniformly expressed in basolateral membranes of normal epithelium, changing to a granular cytoplasmatic expression pattern in carcinomas. Syndecan-2 was observed mainly in a cytoplasmatic granular pattern, with high immunostaining intensity in areas of low Gleason score. E-cadherin was detected in basolateral membrane of normal epithelia showing decreased expression in high Gleason score samples. beta-Catenin was found in cell membranes of normal epithelia changing its distribution toward the nucleus and cytoplasm in carcinoma samples. We concluded that changes in expression and cell distribution of E-cadherin and beta-catenin correlated with the progression degree of prostate adenocarcinoma, suggesting a role of these molecules as markers of progression and prognosis. Furthermore, changes in the pattern expression of syndecan-1 and -2 indicate that both molecules may be involved in the EMT and tumor progression of prostate cancer.

Expression of Multidrug Resistance Proteins in Prostate Cancer Is Related With Cell Sensitivity to Chemotherapeutic Drugs

Multidrug resistance (MDR) proteins have been associated with the lack of chemotherapy response. Expression of these proteins has been described in the prostate, but there is no information about their role in the chemotherapy response of prostate cancer (PC). We studied the gene and protein expression of MDR proteins in primary cell cultures from PC tumors and PC cell lines, their relationship with chemotherapy and their effects on cell survival.

Nonlinear Time Series Analysis by Using Gamma Growing Neural Gas

In this paper, we investigate the properties of the Gamma Growing Neural Gas (γ-GNG) model for the analysis of nonlinear time series. This model includes a temporal context descriptor based on a short term memory structure called Gamma memory. It is shown that γ-GNG can approximately reconstruct the space-state, and filter out additive noise. Simulation results on two data sets are presented: Lorenz system and NH3-FIR Laser time series.

Evaluation of MENT on primary cell cultures from benign prostatic hyperplasia and prostate carcinoma.

7-alpha-Methyl-19-Nortestosterone (MENT) is a synthetic androgen more potent than testosterone (T) and cannot be reduced at 5-alpha position. No important effects of MENT on prostate growth have been reported. However, little is known about the effect of MENT on benign prostatic hyperplasia (BPH) or prostate carcinoma (CaP). We evaluate the effect of MENT, T and dihydrotestosterone (DHT) on secretion, proliferation and gene expression of primary cell cultures from human BPH and CaP. Moreover, the effect of these androgens was examined in the presence of finasteride to determine the influence of the 5-alpha reductase (5-AR) activity on the androgenic potency. BPH and CaP primary cultures were treated with 0, 1, 10 and 100 nM of T, MENT or DHT during 24 and 48 h. Prostate-specific antigen (PSA) was measured by micro particles immunoassay and proliferation rate by spectrophotometric assay (MTT) and by the immunochemical detection of the proliferation marker Ki-67. Gene expression of FGF8b (androgen sensitive gene) was evaluated by semi-quantitative RT-PCR. Results showed that MENT treatments increased PSA secretion and proliferation rate with a potency ranged between T and DHT. Similar effects of MENT were observed in both BPH and CaP cultures. The studies with finasteride showed that in BPH and CaP cells, the conversion of T into DHT significantly contributes to its effect on the proliferation and PSA secretion, and corroborated the resistance of MENT to the 5-AR. The effect of MENT on the gene expression of FGF8b in CaP cells was similar to T and lower than DHT. It is concluded that MENT increases proliferative and secretory activities and gene expression on pathological prostate cells although in less extent than the active metabolite DHT. Furthermore, the fall of endogenous concentration of T during MENT treatment anticipates that this androgen will be of low impact for the prostate.

Segment Growing Neural Gas for Nonlinear Time Series Analysis

In this work we propose an extension to Growing Neural Gas (GNG) for dealing with the spatiotemporal quantization of time series. The two main changes to the original GNG algorithm are the following. First, the basic unit of the GNG network is changed from a node to a linear segment joining two nodes. Secondly, temporal connections between neighboring units in time are added. The proposed algorithm called Segment GNG (SGNG) is compared with the original GNG and Merge GNG algorithms using three benchmark time series: Rossler, Mackey-Glass and \(\text {NH}_{3}\) Laser. The algorithms are applied to the quantization of trajectories in the state space representation of these time series. The results show that the SGNG outperforms both GNG and Merge GNG in terms of quantization error and temporal quantization error.

A strategy for time series prediction using Segment Growing Neural Gas

Segment Growing Neural Gas (Segment-GNG) has been recently proposed as a new spatiotemporal quantization method for time series. Unlike traditional quantization algorithms that are prototype-based, Segment-GNG uses segments as basic units of quantization. In this paper we extend the Segment-GNG model in order to deal with time series prediction. First Segment-GNG makes a quantization of the trajectories in the state-space representation of the time series. Then a local prediction model is associated with each segment, which allows us to make predictions. The proposed model is tested with the Mackey-Glass and Lorenz chaotic time series in one-step ahead prediction tasks. The results obtained are competitive with the best results published in the literature.

Unconfined Aquifer Permeability near hand-dug Wells in the Coastal and Interior dryland of the Libertador General Bernardo O'Higgins Region, Chile

In the dryland of the Libertador General Bernardo O?Higgins Region in Chile, most farmers rely on wide and shallow hand-dug wells as their primary source of water during the dry summer. Few of these wells have sufficient yield for more than domestic use (human consumption, livestock, and irrigation of a subsistence garden). To more accurately assess available groundwater resources, saturated hydraulic conductivity (K) of aquifers in the eight counties that comprise this region?s dryland was estimated using evacuation and recovery tests in 353 hand-dug wells. K followed a log-normal distribution and ranged over nearly five orders of magnitude. County median K varied by a factor of 5 and a slight increasing K trend in the southward direction. In one northern county, less than 4% of the sites had K > 4 m d-1. In the t two southernmost counties, approximately one-quarter of the K values exceeded 4 m d-1. This is approximately the minimum K required for a typical well (1 m diameter and 3 m depth below the water table) to yield 1 L s-1 of water, which is roughly the yield required to irrigate 1 ha. Aquifers located where parent material was predominantly intrusive granite had slightly higher and statistically significant K than those formed predominantly of metamorphic and sedimentary rock. A semi-variogram of K provided weak evidence of a characteristic length scale of approximately 4 km.