蛋白顺序性抗原决定簇的多参数综合预测

THE MULTI-PARAMETER PREDICTION OF POTENTIAL PROTEIN ANTIGENIC DETERMINANTS

通过分析氨基酸序列的某些参数，可达到预测抗原决定簇。但没有一种单参数预测获得较满意结果，故有必要进行多参数联立综合预测。选用６个参数加权叠加。首先假设形成表位的残基上的原子数正比于Ｎ２／３（Ｎ为整蛋白中残基原子数），再对３４个共含１６９个已知抗原位点的蛋白进行训练性综合预测，选出最佳权重。结果表明，Ｈｏｐｐ＆Ｗｏｏｄｓ亲水性参数和Ｊａｎｉｎ可及性参数在综合中很重要。对ＣＲＧＰ，ＳＣＸ２－ＡＮＤＡＮ，ＴＮＦＡ－ＨＵＭＡＮ，ＳＯＭＡ－ＢＯＶＩＮ，ＨＳ７Ｈ－ＨＵＭＡＮ这５个已知位点蛋白进行回顾性分析，初次预测率为４５．８％，再配以β－转角结构，二次预测率为４７．４％，结果可提供实验参考。

There are many methods and paremeters for predicting protein antigenic determinants by analysing amino acid sequences.Some of these are successful.Nevertheless,no single parameter set is able to predict all antigenic determinants.It is necessary to combine various parameters to generate a composite surface profile for prediction.Six common and efficient parameters are used in these search presented here,including Hopp & Woods hydrophilicity,HPL Chydrophilicity(Parker),chargedistribution,accessibility(Janin),mobility(Karplus & Schulz),antigenicity(Welling).βturns are also considered besides them.Linear model is evaluated by a statistical method.169 experimentally determined epitope segments in 34 proteins are trained to obtain the optimal set of weights.Each target protein is assumed to be globelike.Only residues on its shell form epitopes.So the atom number of the epitope residues is in proportional to N2/3(N for the total number of atoms of residues in the intact protein).The train result indicates that Hopp & Woods hydrophilicity and Janin accessibility parameters play important roles in the composite prediction.The method has preliminary average accuracy of about 40%.Five proteins are tested,which a repredicted about 24 epitopes.And get a result to support the method.