Fukunishi Yoshifumi, Nakamura Haruki
Biological Information Research Center (BIRC), National Institute of Advanced Industrial Science and Technology (AIST), 2-41-6, Aomi, Koto-ku, Tokyo 135-0064, Japan.
J Chem Inf Model. 2008 Jan;48(1):148-56. doi: 10.1021/ci700306s. Epub 2008 Jan 1.
The low accuracy of predicted docking scores is critical at in silico drug screening. In order to improve the accuracy of docking scores, we approximated the protein-compound binding free energy as a linear combination of the raw docking scores of a target compound with many different protein pockets. The coefficients of the linear combination were estimated by the similarities among proteins, simply by using the amino-acid sequence similarities or identities of the proteins. This method was applied to in silico screening of the active compounds of five target proteins, and it increased the hit ratio by approximately four to five times compared to that given only by the raw docking scores in every case. The hit ratio also became robust against differences of target proteins.
在计算机辅助药物筛选中，预测对接分数的低准确性是至关重要的。为了提高对接分数的准确性，我们将蛋白质-化合物结合自由能近似为目标化合物与许多不同蛋白质口袋的原始对接分数的线性组合。线性组合的系数通过蛋白质之间的相似性来估计，简单地使用蛋白质的氨基酸序列相似性或同一性。该方法应用于五种目标蛋白活性化合物的计算机辅助筛选，与仅使用原始对接分数的情况相比，在每种情况下命中率提高了约四到五倍。命中率也变得对目标蛋白的差异具有鲁棒性。
