基于改进的BP神经网络对连续B细胞表位的预测
发布时间：2018-08-11 20:41
【摘要】：随着人类基因工程的高效进行,基因组测序数据快速的增加,产生了一门新兴交叉学科——生物信息学。在生物信息学中人们不可能对所有的生物学数据进行实验验证,为了能进行更加有效地实验,充分利用有限的实验资源,对这些生物数据进行分析、整理和有效地预测就显得十分的重要。 B淋巴细胞是人体内十分重要的免疫细胞,其分化成熟于骨髓,进而在Th细胞辅助下,在外周的淋巴组织内与抗原特异性结合,进一步分化为浆细胞,从而分泌抗体,进行免疫活动。B细胞成熟的重要标志就是细胞膜表达了免疫球蛋白IgM、IgD和Igα/Igβ链构成的B细胞受体(BCR)。免疫球蛋白IgM、IgD能特异性的识别抗原,与抗原结合后,再通过电信号形式将信息传递给Igα/Igβ链,由Igα/Igβ链把信号传递到细胞内部,促使B细胞的进一步分化,实现免疫应答。B细胞只有识别抗原(antigen)之后,在抗原的刺激作用下,才能启动免疫应答,免疫作用得以发挥。由此可知,抗原在免疫系统中起着十分重要的作用。抗原是一段蛋白质片段,它在免疫应答中能够与免疫细胞受体结合,在免疫应答中起关键作用。通常将能与B淋巴细胞特异性结合的抗原,称为B细胞表位。由此可见对B细胞表位的预测是特别的重要。 B细胞表位分为连续表位和不连续表位,对于不连续表位的预测需要确定抗原的空间三维结构,因此存在着很大的困难,目前国际上多数都是对连续B细胞表位进行理论筛选。为了对连续B细胞表位做出快速有效地初步理论筛选,提高鉴定实验的成功率,本文应用改进的BP神经网络进行连续B细胞表位理论预测研究,并最终建立了B细胞表位的预测模型,与国内外现有的同类预测模型相比,本模型具有更为优越的预测表现(AUC=0.723)。为了进一步验证模型的性能,本文应用建立的模型对环子孢子蛋白进行了预测,取得了更为满意的效果。
[Abstract]:With the high efficiency of human genetic engineering, genome sequencing data increase rapidly, which has produced a new cross-discipline-bioinformatics. In bioinformatics, it is impossible to verify all the biological data. In order to be able to experiment more effectively and make full use of the limited experimental resources, we can analyze these biological data. B lymphocytes are the most important immune cells in the human body, which differentiate and mature in bone marrow, and then are assisted by Th cells. In peripheral lymphoid tissues, antigen-specific binding, further differentiation into plasma cells, secretion of antibodies, and maturation of immunoglobulin IgMN IgD and Ig 伪 / Ig 尾 chain B cell receptor (BCR). Is an important marker for the maturation of immunoglobulin. The immunoglobulin IgMU IgD can specifically recognize the antigen, bind to the antigen, then transmit the information to the Ig 伪 / Ig 尾 chain in the form of electrical signals, and the signal is transmitted to the cell interior by the Ig 伪 / Ig 尾 chain, which promotes the further differentiation of the B cell. The immune response. B cells can initiate the immune response only after the antigen (antigen) has been recognized and the immune response can be brought into play under the stimulation of the antigen. Therefore, antigens play a very important role in the immune system. Antigen is a fragment of protein that binds to immune cell receptors in immune responses and plays a key role in immune responses. B-cell epitopes are commonly referred to as B-cell epitopes that specifically bind to B lymphocytes. Therefore, the prediction of B cell epitopes is particularly important. B cell epitopes are divided into continuous epitopes and discontinuous epitopes. For the prediction of discontinuous epitopes, it is necessary to determine the three-dimensional structure of antigens. Therefore, there are great difficulties. At present, most of the continuous B cell epitopes are screened theoretically. In order to make rapid and effective theoretical screening of continuous B cell epitopes and improve the success rate of identification experiments, the improved BP neural network is used to predict the continuous B cell epitopes. Finally, the prediction model of B cell epitopes is established. Compared with the existing prediction models, this model has more superior prediction performance (AUC=0.723). In order to further verify the performance of the model, the model has been used to predict the ring sporozoite protein, and more satisfactory results have been obtained.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：R392
本文编号：2178180
[Abstract]:With the high efficiency of human genetic engineering, genome sequencing data increase rapidly, which has produced a new cross-discipline-bioinformatics. In bioinformatics, it is impossible to verify all the biological data. In order to be able to experiment more effectively and make full use of the limited experimental resources, we can analyze these biological data. B lymphocytes are the most important immune cells in the human body, which differentiate and mature in bone marrow, and then are assisted by Th cells. In peripheral lymphoid tissues, antigen-specific binding, further differentiation into plasma cells, secretion of antibodies, and maturation of immunoglobulin IgMN IgD and Ig 伪 / Ig 尾 chain B cell receptor (BCR). Is an important marker for the maturation of immunoglobulin. The immunoglobulin IgMU IgD can specifically recognize the antigen, bind to the antigen, then transmit the information to the Ig 伪 / Ig 尾 chain in the form of electrical signals, and the signal is transmitted to the cell interior by the Ig 伪 / Ig 尾 chain, which promotes the further differentiation of the B cell. The immune response. B cells can initiate the immune response only after the antigen (antigen) has been recognized and the immune response can be brought into play under the stimulation of the antigen. Therefore, antigens play a very important role in the immune system. Antigen is a fragment of protein that binds to immune cell receptors in immune responses and plays a key role in immune responses. B-cell epitopes are commonly referred to as B-cell epitopes that specifically bind to B lymphocytes. Therefore, the prediction of B cell epitopes is particularly important. B cell epitopes are divided into continuous epitopes and discontinuous epitopes. For the prediction of discontinuous epitopes, it is necessary to determine the three-dimensional structure of antigens. Therefore, there are great difficulties. At present, most of the continuous B cell epitopes are screened theoretically. In order to make rapid and effective theoretical screening of continuous B cell epitopes and improve the success rate of identification experiments, the improved BP neural network is used to predict the continuous B cell epitopes. Finally, the prediction model of B cell epitopes is established. Compared with the existing prediction models, this model has more superior prediction performance (AUC=0.723). In order to further verify the performance of the model, the model has been used to predict the ring sporozoite protein, and more satisfactory results have been obtained.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：R392
【参考文献】
相关期刊论文 前1条
1 蒋良孝 ,李超群;基于BP神经网络的函数逼近方法及其MATLAB实现[J];微型机与应用;2004年01期
，本文编号：2178180
本文链接：https://www.wllwen.com/xiyixuelunwen/2178180.html
最近更新
教材专著
