中文摘要：

对蛋白质结构预测问题进行了描述，根据蛋白质结构预测问题在三维欧氏空间的连续模型，通过拟物策略找到了相应的数学模型．由于引入了弹性势能和嵌入势能，将一个有约束的问题转化为一个无约束的问题，并在梯度下降法求解的基础上，设计出一种变步长梯度下降的求解方法．通过实例检测，变步长梯度下降法比梯度下降法大大节省了计算时间，且所得结果的能量比梯度下降法所得结果的能量更低．

英文摘要：

The problem of the protein structure prediction was described. According to the continuous model of the 3D Euclidean space for protein structure prediction which is proposed in the international. A related arithmetic model was found through quasiphysical ways. As the elasticity potential energy and the press potential energy are leaded into, so that a bondage problem is transformed into a nonbondage problem, and a resolvent of grads-decline by advanced step is formulated on the basis of the resolvent of grads-decline. In all test cases, it was far more fast than the resolvent of grads-decline and it＇s energy of results was far more low than the resolvent of grads-decline.