Machine learning algorithms for spatial data. Case studies: Environmental pollution, natural hazards, renewable resources

This paper presents general problems and approaches for the spatial data analysis using machine learning algorithms. Machine learning is a very powerful approach to adaptive data analysis, modelling and visualisation. The key feature of the machine learning algorithms is that they learn from empirical data and can be used in cases when the modelled environmental phenomena are hidden, nonlinear, noisy and highly variable in space and in time. Most of the machines learning algorithms are universal and adaptive modelling tools developed to solve basic problems of learning from data: classification/pattern recognition, regression/mapping and probability density modelling. In the present report some of the widely used machine learning algorithms, namely artificial neural networks (ANN) of different architectures and Support Vector Machines (SVM), are adapted to the problems of the analysis and modelling of geo-spatial data. Machine learning algorithms have an important advantage over traditional models of spatial statistics when problems are considered in a high dimensional geo-feature spaces, when the dimension of space exceeds 5. Such features are usually generated, for example, from digital elevation models, remote sensing images, etc. An important extension of models concerns considering of real space constrains like geomorphology, networks, and other natural structures. Recent developments in semi-supervised learning can improve modelling of environmental phenomena taking into account on geo-manifolds. An important part of the study deals with the analysis of relevant variables and models' inputs. This problem is approached by using different feature selection/feature extraction nonlinear tools. To demonstrate the application of machine learning algorithms several interesting case studies are considered: digital soil mapping using SVM, automatic mapping of soil and water system pollution using ANN; natural hazards risk analysis (avalanches, landslides), assessments of renewable resources (wind fields) with SVM and ANN models, etc. The dimensionality of spaces considered varies from 2 to more than 30. Figures 1, 2, 3 demonstrate some results of the studies and their outputs. Finally, the results of environmental mapping are discussed and compared with traditional models of geostatistics.

Fine-scale spatial genetic structure and gene dispersal in Silene latifolia.

Plants are sessile organisms, often characterized by limited dispersal. Seeds and pollen are the critical stages for gene flow. Here we investigate spatial genetic structure, gene dispersal and the relative contribution of pollen vs seed in the movement of genes in a stable metapopulation of the white campion Silene latifolia within its native range. This short-lived perennial plant is dioecious, has gravity-dispersed seeds and moth-mediated pollination. Direct measures of pollen dispersal suggested that large populations receive more pollen than small isolated populations and that most gene flow occurs within tens of meters. However, these studies were performed in the newly colonized range (North America) where the specialist pollinator is absent. In the native range (Europe), gene dispersal could fall on a different spatial scale. We genotyped 258 individuals from large and small (15) subpopulations along a 60 km, elongated metapopulation in Europe using six highly variable microsatellite markers, two X-linked and four autosomal. We found substantial genetic differentiation among subpopulations (global F(ST)=0.11) and a general pattern of isolation by distance over the whole sampled area. Spatial autocorrelation revealed high relatedness among neighboring individuals over hundreds of meters. Estimates of gene dispersal revealed gene flow at the scale of tens of meters (5-30 m), similar to the newly colonized range. Contrary to expectations, estimates of dispersal based on X and autosomal markers showed very similar ranges, suggesting similar levels of pollen and seed dispersal. This may be explained by stochastic events of extensive seed dispersal in this area and limited pollen dispersal.

Fast subplasma membrane Ca2+ transients control exo-endocytosis of synaptic-like microvesicles in astrocytes

Astrocytes are the most abundant glial cell type in the brain. Although not apposite for long-range rapid electrical communication, astrocytes share with neurons the capacity of chemical signaling via Ca(2+)-dependent transmitter exocytosis. Despite this recent finding, little is known about the specific properties of regulated secretion and vesicle recycling in astrocytes. Important differences may exist with the neuronal exocytosis, starting from the fact that stimulus-secretion coupling in astrocytes is voltage independent, mediated by G-protein-coupled receptors and the release of Ca(2+) from internal stores. Elucidating the spatiotemporal properties of astrocytic exo-endocytosis is, therefore, of primary importance for understanding the mode of communication of these cells and their role in brain signaling. We here take advantage of fluorescent tools recently developed for studying recycling of glutamatergic vesicles at synapses (Voglmaier et al., 2006; Balaji and Ryan, 2007); we combine epifluorescence and total internal reflection fluorescence imaging to investigate with unprecedented temporal and spatial resolution, the stimulus-secretion coupling underlying exo-endocytosis of glutamatergic synaptic-like microvesicles (SLMVs) in astrocytes. Our main findings indicate that (1) exo-endocytosis in astrocytes proceeds with a time course on the millisecond time scale (tau(exocytosis) = 0.24 +/- 0.017 s; tau(endocytosis) = 0.26 +/- 0.03 s) and (2) exocytosis is controlled by local Ca(2+) microdomains. We identified submicrometer cytosolic compartments delimited by endoplasmic reticulum tubuli reaching beneath the plasma membrane and containing SLMVs at which fast (time-to-peak, approximately 50 ms) Ca(2+) events occurred in precise spatial-temporal correlation with exocytic fusion events. Overall, the above characteristics of transmitter exocytosis from astrocytes support a role of this process in fast synaptic modulation.

Development of allocentric spatial memory abilities in children from 18 months to 5 years of age.

Episodic memories for autobiographical events that happen in unique spatiotemporal contexts are central to defining who we are. Yet, before 2 years of age, children are unable to form or store episodic memories for recall later in life, a phenomenon known as infantile amnesia. Here, we studied the development of allocentric spatial memory, a fundamental component of episodic memory, in two versions of a real-world memory task requiring 18 month- to 5-year-old children to search for rewards hidden beneath cups distributed in an open-field arena. Whereas children 25-42-months-old were not capable of discriminating three reward locations among 18 possible locations in absence of local cues marking these locations, children older than 43 months found the reward locations reliably. These results support previous findings suggesting that allocentric spatial memory, if present, is only rudimentary in children under 3.5 years of age. However, when tested with only one reward location among four possible locations, children 25-39-months-old found the reward reliably in absence of local cues, whereas 18-23-month-olds did not. Our findings thus show that the ability to form a basic allocentric representation of the environment is present by 2 years of age, and its emergence coincides temporally with the offset of infantile amnesia. However, the ability of children to distinguish and remember closely related spatial locations improves from 2 to 3.5 years of age, a developmental period marked by persistent deficits in long-term episodic memory known as childhood amnesia. These findings support the hypothesis that the differential maturation of distinct hippocampal circuits contributes to the emergence of specific memory processes during early childhood.

Cytostatic lung perfusion results in heterogeneous spatial regional blood flow and drug distribution: evaluation of different cytostatic lung perfusion techniques in a porcine model.

OBJECTIVES: Comparison of doxorubicin uptake, leakage and spatial regional blood flow, and drug distribution was made for antegrade, retrograde, combined antegrade and retrograde isolated lung perfusion, and pulmonary artery infusion by endovascular inflow occlusion (blood flow occlusion), as opposed to intravenous administration in a porcine model. METHODS: White pigs underwent single-pass lung perfusion with doxorubicin (320 mug/mL), labeled 99mTc-microspheres, and Indian ink. Visual assessment of the ink distribution and perfusion scintigraphy of the perfused lung was performed. 99mTc activity and doxorubicin levels were measured by gamma counting and high-performance liquid chromatography on 15 tissue samples from each perfused lung at predetermined localizations. RESULTS: Overall doxorubicin uptake in the perfused lung was significantly higher (P = .001) and the plasma concentration was significantly lower (P < .0001) after all isolated lung perfusion techniques, compared with intravenous administration, without differences between them. Pulmonary artery infusion (blood flow occlusion) showed an equally high doxorubicin uptake in the perfused lung but a higher systemic leakage than surgical isolated lung perfusion (P < .0001). The geometric coefficients of variation of the doxorubicin lung tissue levels were 175%, 279%, 226%, and 151% for antegrade, retrograde, combined antegrade and retrograde isolated lung perfusion, and pulmonary artery infusion by endovascular inflow occlusion (blood flow occlusion), respectively, compared with 51% for intravenous administration (P = .09). 99mTc activity measurements of the samples paralleled the doxorubicin level measurements, indicating a trend to a more heterogeneous spatial regional blood flow and drug distribution after isolated lung perfusion and blood flow occlusion compared with intravenous administration. CONCLUSIONS: Cytostatic lung perfusion results in a high overall doxorubicin uptake, which is, however, heterogeneously distributed within the perfused lung.

The feasibility of 350 μm spatial resolution coronary magnetic resonance angiography at 3 T in humans.

PURPOSE: The purposes of this study were to (1) develop a high-resolution 3-T magnetic resonance angiography (MRA) technique with an in-plane resolution approximate to that of multidetector coronary computed tomography (MDCT) and a voxel size of 0.35 × 0.35 × 1.5 mm³ and to (2) investigate the image quality of this technique in healthy participants and preliminarily in patients with known coronary artery disease (CAD). MATERIALS AND METHODS: A 3-T coronary MRA technique optimized for an image acquisition voxel as small as 0.35 × 0.35 × 1.5 mm³ (high-resolution coronary MRA [HRC]) was implemented and the coronary arteries of 22 participants were imaged. These included 11 healthy participants (average age, 28.5 years; 5 men) and 11 participants with CAD (average age, 52.9 years; 5 women) as identified on MDCT. In addition, the 11 healthy participants were imaged using a method with a more common spatial resolution of 0.7 × 1 × 3 mm³ (regular-resolution coronary MRA [RRC]). Qualitative and quantitative comparisons were made between the 2 MRA techniques. RESULTS: Normal vessels and CAD lesions were successfully depicted at 350 × 350 μm² in-plane resolution with adequate signal-to-noise ratio (SNR) and contrast-to-noise ratio. The CAD findings were consistent among MDCT and HRC. The HRC showed a 47% improvement in sharpness despite a reduction in SNR (by 72%) and in contrast-to-noise ratio (by 86%) compared with the regular-resolution coronary MRA. CONCLUSION: This study, as a first step toward substantial improvement in the resolution of coronary MRA, demonstrates the feasibility of obtaining at 3 T a spatial resolution that approximates that of MDCT. The acquisition in-plane pixel dimensions are as small as 350 × 350 μm² with a 1.5-mm slice thickness. Although SNR is lower, the images have improved sharpness, resulting in image quality that allows qualitative identification of disease sites on MRA consistent with MDCT.

Antibiotic-dependent correlation between drug-induced killing and loss of luminescence in Streptococcus gordonii expressing luciferase.

Measuring antibiotic-induced killing relies on time-consuming biological tests. The firefly luciferase gene (luc) was successfully used as a reporter gene to assess antibiotic efficacy rapidly in slow-growing Mycobacterium tuberculosis. We tested whether luc expression could also provide a rapid evaluation of bactericidal drugs in Streptococcus gordonii. The suicide vectors pFW5luc and a modified version of pJDC9 carrying a promoterless luc gene were used to construct transcriptional-fusion mutants. One mutant susceptible to penicillin-induced killing (LMI2) and three penicillin-tolerant derivatives (LMI103, LMI104, and LMI105) producing luciferase under independent streptococcal promoters were tested. The correlation between antibiotic-induced killing and luminescence was determined with mechanistically unrelated drugs. Chloramphenicol (20 times the MIC) inhibited bacterial growth. In parallel, luciferase stopped increasing and remained stable, as determined by luminescence and Western blots. Ciprofloxacin (200 times the MIC) rapidly killed 1.5 log10 CFU/ml in 2-4 hr. Luminescence decreased simultaneously by 10-fold. In contrast, penicillin (200 times the MIC) gave discordant results. Although killing was slow (&lt; or = 0.5 log10 CFU/ml in 2 hr), luminescence dropped abruptly by 50-100-times in the same time. Inactivating penicillin with penicillinase restored luminescence, irrespective of viable counts. This was not due to altered luciferase expression or stability, suggesting some kind of post-translational modification. Luciferase shares homology with aminoacyl-tRNA synthetase and acyl-CoA ligase, which might be regulated by macromolecule synthesis and hence affected in penicillin-inhibited cells. Because of resemblance, luciferase might be down-regulated simultaneously. Luminescence cannot be universally used to predict antibiotic-induced killing. Thus, introducing reporter enzymes sharing mechanistic similarities with normal metabolic reactions might reveal other effects than those expected.

Telomerase activity and human telomerase reverse transcriptase mRNA expression in soft tissue tumors: correlation with grade, histology, and proliferative activity.

Telomerase activity (TA) is detected in most human cancers but, with few exceptions, not in normal somatic cells. Little is known about TA in soft tissue tumors. We have examined a series of benign and malignant soft tissue tumors for TA using the telomerase repeat amplification protocol assay. Analysis of the expression of the human telomerase reverse transcriptase was also carried out using RT-PCR. TA was undetectable in benign lesions (15 of 15) and low-grade sarcomas (6 of 6) and was detectable in 50% (19 of 38) of intermediate-/high-grade sarcomas. Although the presence of TA in soft tissue tumors is synonymous with malignancy, it is neither a reliable method in making the distinction between reactive/benign and malignant (especially low-grade) lesions nor a reliable marker of tumor aggressiveness. Leiomyosarcomas and storiform/pleomorphic malignant fibrous histiocytomas rarely showed TA, irrespective of their grade. A strong correlation between human telomerase reverse transcriptase mRNA expression and TA was observed, supporting the close relationship between both parameters. No significant relationship was observed between proliferative activity (as assessed by MIB-1 immunolabeling) and TA. We verified that the absence of telomerase expression was not due to the presence of telomerase inhibitors and therefore alternative mechanism(s) for cell immortalization, yet to be determined, seem to be involved in the development and/or maintenance of some soft tissue sarcomas.