The purpose of a Neural Network is to learn data patterns and to recognize them in test data. The neural network is first taught different patterns and is then given a test pattern to determine which of the known patterns the test pattern most resembles. The simplest neural network is made up of three layers: input, hidden, and output. Our focus is the hidden layer. It is made up of connecting weights that enables it to examine the test data and recognize previously learned patterns within the test data. It is then able to display whether the test data contains any of these previously learned patterns. The weights can then be adjusted by a combination of the sigmoid function and the delta function. The data that was acquired by Dr. Joplin, Dr. Moore, and Georgianna has a wide range and two separate channels and each channel has intensity and a background. The first channel is for genes present in up regulated and the second for down regulated genes. The data contains two channels and the network can only deal with one so there is a need for the data to be transformed so that the first and the second channel can be inputted and worked with at the same time. One way to modify the data for use is to divide channel one by channel two. In doing so, genes that are present in both diapause and nondiapause are around one. Genes that should be up regulated are greater than 1 and genes that should be down regulated are below one. A paper that we are going to mimic started with a single layer neural network that was first taught the data. After the neural network was taught the data then genes that had weights and alphas above or below a certain critical value remained within the network while those within the range were eliminated from the network. This process was repeated until only about twenty or forty genes were left. The network maintained accuracy even after so many genes had been eliminated. We are going to duplicate this procedure and try to eliminate some of the seventeen thousand genes so that we will have a more clear idea of what to focus on with the network.
My contribution to this internship is the physical verification of the diapause status, up or down regulated, of Sarcophaga crassipalpis genes. For S. crassipalpis, the diapause state is almost a complete cessation in development and is induced by increasing the hours of uninterrupted night from nine to twelve. Pupae that enter diapause exhibit a variety of characteristics that non-diapausing pupae do not. The alternate is also true. I am working with Dr. Joplin and Dr. Miller to determine which genes are expressed in which state. Our experimental data will then be used to train a neural network to recognize patterns in raw microarray data. We choose our candidate genes for study based on log values assigned during normalization of a heterologous microarray. This microarray was conducted using Drosophila melanogaster DNA and S. crassipalpis RNA and suggests genes that may be differentially regulated. After choosing candidate genes, I will be performing reverse transcription (RT) to create cDNA from RNA isolated from diapausing and nondiapausing pupae. The cDNA will be amplified using polymerase chain reaction (PCR). After PCR, I will perform agarose gel electrophoresis to verify the presence or absence of a product. Presence of a band in the gel for the state indicated by the microarray, with the absence of a band in the opposite state, verifies that the microarray data for that gene is accurate. Once verified, the gene will be used to train a neural network that is being designed by Dr. J. Knisley and Patricia Carey.
|
|
|
|
|
clip 1 clip 2 clip 3 clip 4 clip 5