一个应用于MHCⅡ类分子亲和肽预测的集成分类器

摘要

Classification is a classical issue in the data mining and machine learning field. It produces classifiers that are widely and concretely applied to a lot of fields. As an interdisciplinary subject prospered in recent years, Bioinformatics is sure to become one of the key application trends of classifiers. For such an important issue as MHC class Ⅱ molecule peptides prediction in immunology bioinformatics, there are quite a lot of mature classifiers, each bearing its respective strengths and weaknesses. Therefore, it seems essential to design and realize a classifier that ensembles strengths of different classifiers with stable performances to cope with datasets that are as all-rounded as possible. By ensemble learning method the article designs and realizes a classifier that can be applied to predicting MHC class Ⅱ molecule binding conditions. The system can ensemble an arbitrary number of predictor classifiers in order to achieve a better classification result. Meanwhile, the classifier ensemble learning algorithm can be populated to more fields that are relating to bioinformatics.%分类是数据挖掘和机器学习领域的经典问题,由此产生的分类器目前在许多领域中有着广泛的具体应用.生物信息学作为近几年迅速兴起的交叉学科,自然成为分类器的重点应用方向之一.针对免疫生物信息学中的MHCⅡ类分子亲和肽预测这一重要课题,现阶段已有不少成熟的分类器,但是各分类器表现互有优劣.因而,设计实现一个可以集中各个分类器优势的集成分类器,使其面对尽可能全面的数据集都有稳定表现,就显得十分重要.采用集成学习的方法设计实现一个可以用于预测MHCⅡ类分子绑定情况的分类器,该系统可以将任意数目的预测分类器进行集成,从而达到更好的分类效果.同时,这一分类器集成学习算法可以推广应用到生物信息相关的更广泛领域.