Spelling suggestions: "subject:"enetic approach"" "subject:"benetic approach""
1 |
A genetic approach to identify the requirements for phosphotyrosine specific outputs of Neu/ErbB2Hossain, Noor 04 1900 (has links)
<p> DER, the Drosophila Epidermal Growth Factor Receptor (DEgfr) is
the only known fly orthologue of vertebrate Neu/ErbB2 receptor tyrosine
kinase family. Receptor Tyrosine Kinases (RTKs) like DER and ErbB2 play
an important role in regulating cell differentiation, cell proliferation and cell
survival in metazoan animals. Neu/ErbB2 is over-expressed in 20-30% human
breast cancers, which correlates with poor clinical prognosis in cancer
patients. </p> <p> Our previous studies showed that rat-NeriJErbB2 could successfully
signal in vivo using Drosophila adaptor and second messenger molecules.
Here we regenerated the transgenic fly lines with various neu add-back alleles.
We further re-established mis-expression phenotypes in various adult
structures such as wings and eyes, the tissues known to require DEgfr
signaling. By using genetic approach, we have demonstrated that the tyrosine
residue at the 1028 site (NeuYA), might have an inhibitory role in RTK
signaling. In addition we have already generated a number of double add-back
neu alleles where tyrosine site at the 1028 site (neuYA) was added back to
another Neu allele and made neuYAB, neuYAc neuYAD and neuYAE. Transgenic flies
with these alleles will be generated to further study the inhibitory role of
Neu^YA. </p> <p> Finally, our on going large-scale genetic screening is likely to reveal the
component(s) of NeuYE (Y1253) pathway that does not utilize the function of
Ras. </p> / Thesis / Master of Science (MSc)
|
2 |
Optimizing neural network structures: faster speed, smaller size, less tuningLi, Zhe 01 January 2018 (has links)
Deep neural networks have achieved tremendous success in many domains (e.g., computer vision~\cite{Alexnet12,vggnet15,fastrcnn15}, speech recognition~\cite{hinton2012deep,dahl2012context}, natural language processing~\cite{dahl2012context,collobert2011natural}, games~\cite{silver2017mastering,silver2016mastering}), however, there are still many challenges in deep learning comunity such as how to speed up training large deep neural networks, how to compress large nerual networks for mobile/embed device without performance loss, how to automatically design the optimal network structures for a certain task, and how to further design the optimal networks with improved performance and certain model size with reduced computation cost.
To speed up training large neural networks, we propose to use multinomial sampling for dropout, i.e., sampling features or neurons according to a multinomial distribution with different probabilities for different features/neurons. To exhibit the optimal dropout probabilities, we analyze the shallow learning with multinomial dropout and establish the risk bound for stochastic optimization. By minimizing a sampling dependent factor in the risk bound, we obtain a distribution-dependent dropout with sampling probabilities dependent on the second order statistics of the data distribution. To tackle the issue of evolving distribution of neurons in deep learning, we propose an efficient adaptive dropout (named evolutional dropout) that computes the sampling probabilities on-the-fly from a mini-batch of examples.
To compress large neural network structures, we propose a simple yet powerful method for compressing the size of deep Convolutional Neural Networks (CNNs) based on parameter binarization. The striking difference from most previous work on parameter binarization/quantization lies at different treatments of $1\times 1$ convolutions and $k\times k$ convolutions ($k>1$), where we only binarize $k\times k$ convolutions into binary patterns. By doing this, we show that previous deep CNNs such as GoogLeNet and Inception-type Nets can be compressed dramatically with marginal drop in performance. Second, in light of the different functionalities of $1\times 1$ (data projection/transformation) and $k\times k$ convolutions (pattern extraction), we propose a new block structure codenamed the pattern residual block that adds transformed feature maps generated by $1\times 1$ convolutions to the pattern feature maps generated by $k\times k$ convolutions, based on which we design a small network with $\sim 1$ million parameters. Combining with our parameter binarization, we achieve better performance on ImageNet than using similar sized networks including recently released Google MobileNets.
To automatically design neural networks, we study how to design a genetic programming approach for optimizing the structure of a CNN for a given task under limited computational resources yet without imposing strong restrictions on the search space. To reduce the computational costs, we propose two general strategies that are observed to be helpful: (i) aggressively selecting strongest individuals for survival and reproduction, and killing weaker individuals at a very early age; (ii) increasing mutation frequency to encourage diversity and faster evolution. The combined strategy with additional optimization techniques allows us to explore a large search space but with affordable computational costs.
To further design the optimal networks with improved performance and certain model size under reduced computation cost, we propose an ecologically inspired genetic approach for neural network structure search , that includes two types of succession: primary and secondary succession as well as accelerated extinction. Specifically, we first use primary succession to rapidly evolve a community of poor initialized neural network structures into a more diverse community, followed by a secondary succession stage for fine-grained searching based on the networks from the primary succession. Accelerated extinction is applied in both stages to reduce computational cost. In addition, we also introduce the gene duplication to further utilize the novel block of layers that appeared in the discovered network structure.
|
3 |
Ecology of Tigers in Churia Habitat and a Non-Invasive Genetic Approach to Tiger Conservation in Terai Arc, NepalThapa, Kanchan 13 October 2014 (has links)
Tigers (Panthera tigris tigris) can be viewed as a proxy for intact and healthy ecosystems. Their wild populations have plummeted to fewer than 3,200 individuals in the last four decades and threats to these apex predators are mounting rather than diminishing. Global conservation bodies (Global Tiger Initiative, World Wildlife Fund, Wildlife Conservation Society, Panthera etc.) have recently called for solidarity and scaling up of conservation efforts to save tigers from extinction.
In South Asia, tiger habitat ranges from tropical evergreen forests, dry arid regions and sub-tropical alluvial floodplains, to temperate mixed deciduous forest. The churia habitat is relatively unstudied and is considered a young and geologically fragile mountain range in Nepal. The contribution of the churia habitat to tiger conservation has not been considered, since modern conservation started in 1970's. This study focuses on the ecology of the tiger with respect to population density, habitat use, and prey occupancy and density, in the churia habitat of Chitwan National Park. This study also includes the first assessment of genetic diversity, genetic structure, and gene flow of tigers across the Terai Arc Landscape- Nepal. The Terai Arc Landscape harbors the only remaining tiger population found across the foothills of the Himalayas in Nepal and northwest India. I used a combination of camera-trapping techniques, which have been a popular and robust method for monitoring tiger populations across the landscape, combined with a noninvasive genetic approach to gain information on tigers, thus adding new information relevant to global tiger conservation.
I investigated tiger, leopard (Panthera pardus fusca), and prey densities, and predicted the tiger density across the Churia habitat in Chitwan National Park. I used a camera-trap grid with 161 locations accumulating 2,097 trap-nights in a 60 day survey period during the winter season of 2010-2011. Additionally, I used distance sampling techniques for estimating prey density in the churia habitat by walking 136 km over 81 different line transects. The team photographed 31 individual tigers and 28 individual leopards along with 25 mammalian species from a sampling area of 536 km² comprising Churia and surrounding areas. Density estimates of tigers and leopards were 2.2 (SE 0.42) tigers and 4.0 (SE 1.00) leopards per 100 km². Prey density was estimated at 62.7 prey animals per 100 km² with contributions from forest ungulates to be 47% (sambar Rusa unicolor, chital Axis axis, barking deer Muntiacus muntjak, and wild pigs Sus scrofa). Churia habitat within Chitwan National Park is capable of supporting 5.86 tigers per 100 km² based on applying models developed to predict tiger density from prey density. My density estimates from camera-traps are lower than that predicted based on prey availability, which indicates that the tiger population may be below the carrying capacity. Nonetheless, the churia habitat supports 9 to 36 tigers, increasing estimates of current population size in Chitwan National Park. Based on my finding, the Churia habitat should no longer remain ignored because it has great potential to harbor tigers. Conservation efforts should focus on reducing human disturbance to boost prey populations to potentially support higher predator numbers in Churia.
I used sign surveys within a rigorous occupancy framework to estimate probability of occupancy for 5 focal prey species of the tiger (gaur Bos gaurus, sambar, chital, wild pig, and barking deer); as well as probability of tiger habitat use within 537 km² of churia habitat in Chitwan National Park. Multi-season, auto-correlation models allowed me to make seasonal (winter versus summer) inferences regarding changes in occupancy or habitat use based on covariates influencing occupancy and detection. Sambar had the greatest spatial distribution across both seasons, occupying 431-437 km² of the churia habitat, while chital had the lowest distribution, occupying only 100-158 km². The gaur population showed the most seasonal variation from 318- 413 km² of area occupied, with changes in occupancy suggesting their migration out of the lowland areas in the summer and into the churia in the winter. Wild pigs showed the opposite, moving into the churia in the summer (444 km² area occupied) and having lower occupancy in the winter (383 km²). Barking deer were widespread in both seasons (329 - 349 km²). Tiger probability of habitat use Ψ SE(Ψ) was only slightly higher in winter 0.63 (SE 0.11) than in summer 0.54 (SE 0.21), but confidence intervals overlapped and area used was very similar across seasons, from 337 - 291 km². Fine-scale variation in tiger habitat use showed that tigers intensively use certain areas more often than others across the seasons. The proportion of available habitat positively influenced occupancy for the majority of prey species and tigers. Human disturbance had a strong negative influence on the distribution of the majority of prey species but was positively related to tiger habitat use. Tigers appear to live in areas with high disturbance, thus increasing the risk of human-tiger conflict in the churia habitat. Thus, efforts to reduce human disturbance would be beneficial to reducing human wildlife conflict, enriching prey populations, and would potentially support more tigers in churia habitat of Nepal. Overall, I found high prey occupancy and tiger habitat use, suggesting that the churia is highly valuable habitat for tigers and should no longer be neglected or forgotten in tiger conservation planning.
Thirdly, I assessed genetic variation, genetic structure, and gene flow of the tigers in the Terai Arc Landscape, Nepal. I opportunistically collected 770 scat samples from 4 protected areas and 5 hypothesized corridors across the Terai Arc Landscape. Historical landuse change in the Terai Arc was extracted from Anthrome data sets to relate landuse change to potential barriers and subsequent hypothesized bottleneck events in the landscape. I used standard genetic metrics (allelic diversity and heterozygosity) to estimate genetic variation in the tiger population. Using program Structure (non-spatial) and TESS (spatial), I defined the putative genetic clusters present in the landscape. Migrant analysis was carried out in Geneclass and Bayesass for estimating contemporary gene flow. I tested for a recent population bottleneck with the heterozygosity test using program Bottleneck. Of the 700 samples, 396 were positive for tiger (57% success). Using an 8 multilocus microsatellite assay, I identified 78 individual tigers. I found large scale landuse changes across the Terai Arc Landscape due to conversion of forest into agriculture in last two centuries and I identified areas of suspected barriers. I found low levels of genetic variation (expected heterozygosity = 0.61) and moderate genetic differentiation (F<sub>ST</sub> = 0.14) across the landscape, indicative of sub-population structure and potential isolation of sub-populations. I detected three genetic clusters across the landscape consistent with three demographic tiger sub-populations occurring in Chitwan-Parsa, Bardia, and Suklaphanta protected areas. I detected 10 migrants across all study sites confirming there is still some dispersal mediated gene flow across the landscape. I found evidence of a bottleneck signature, especially around the lowland forests in the Terai, likely caused by large scale landuse change in last two centuries, which could explain the low levels of genetic variation detected at the sub-population level. These findings are highly relevant to tiger conservation indicating that efforts to protect source sites and to improve connectivity are needed to augment gene flow and genetic diversity across the landscape.
Finally, I compared the abundance and density of tigers obtained using two non-invasive sampling techniques: camera-trapping and fecal DNA sampling. For cameras: I pooled the 2009 camera-trap data from the core tiger population across the lowland areas of Chitwan National Park. I sampled 359 km² of the core area with 187 camera-trap locations spending 2,821 trap-nights of effort. I obtained 264 identifiable photographs and identified a total of 41 individual tigers. For genetics, I sampled 325 km² of the core area along three spatial routes, walking a total of 1,173 km, collecting a total of 420 tiger fecal samples in 2011. I identified 36 tigers using the assay of 8 multilocus genotypes and captured them 42 times. I analyzed both data types separately for estimating density and jointly in an integrated model using both traditional, and spatial, capture-recapture frameworks. Using Program MARK and the model averaged results, my abundance estimates were 46 (SE 1.86) and 44 (SE 9.83) individuals from camera and genetic data, respectively. Density estimates (tigers per 100 km²) via traditional buffer strip methods using half of the Mean Maximum Distance Moved (½ MMDM) as the buffer surrounding survey grids, were 4.01 (SE 0.64) for camera data and 3.49 (SE 1.04) for genetic data. Spatially explicit capture recapture models resulted in lower density estimates both in the likelihood based program DENSITY at 2.55 (SE 0.59) for camera-trap data and 2.57 (SE 0.88) for genetic data, while the Bayesian based program SPACECAP estimates were 2.44 (SE 0.30) for camera-trap data and 2.23 (SE 0.46) for genetic data. Using a spatially explicit, integrated model that combines data from both cameras and genetics, density estimates were 1.47 (SD 0.20) tigers per 100 km² for camera-trap data and 1.89 (SD 0.36) tigers per 100 km² for genetic data. I found that the addition of camera-trap data improved precision in genetic capture-recapture estimates, but not visa-versa, likely due to low numbers of recaptures in the genetic data. While a non-invasive genetic approach can be used as a stand-alone capture-recapture method, it may be necessary to increase sample size to obtain more recaptures. Camera-trap data may provide a more precise estimates, but genetic data returns more information on other aspect of genetic health and connectivity. Combining data sets in an integrated modeling framework, aiding in pinpointing strengths and weaknesses in data sets, thus ultimately improving modeling inference. / Ph. D.
|
Page generated in 0.0632 seconds