CHDOCK: a hierarchical docking approach for(2)
【作者】网站采编
【关键词】
【摘要】:Similar trends can also be observed in the results for unbound docking,though the performance differences among different algorithms are not as much as those for bound docking due to the impact of conSimilar trends can also be observed in the results for unbound docking,though the performance differences among different algorithms are not as much as those for bound docking due to the impact of conformational changes in the unbound ,CHDOCK also performed significantly better than the other three methods for unbound docking and obtained a success rate of 30.66%,44.81%,and 68.40% for top 1,10,and 100 predictions,respectively,in comparison to those of 26.42%,36.79%,and 66.51% for M-ZDOCK,19.34%,31.60%,and 63.68%for SAM,and 11.79%,30.66%,and 58.49% for SymmDock (Table 1 and Fig.1).
Besides the success rate of docking,we have also compared the average root mean square deviation(RMSD) of ‘hit(s)’ (i.e.,successful binding mode predictions) for both bound docking and unbound docking with the other three programs when the top 1,10,100 predictions were results are listed in Table 2 and the corresponding results are shown in Table 2 and Fig.2,we can see that CHDOCK also performed much better and obtained moreaccurate binding modes than the other three programs for both bound docking and unbound bound docking,CHDOCK obtained an average RMSD of 1.10,1.51 and 2.11 ? for top 1,10 and 100 predictions,respectively,in comparison to those of 2.27,2.80 and 3.46 ? for the second-best method for unbound docking,similar results can also be obtained an average RMSD of 2.54,3.12 and 4.07 ? for top 1,10,100 predictions,respectively,while M-ZDOCK obtained a higher RMSD of 3.26,3.86 and 5.07 ?.Interestingly,one can also note that among the four docking programs,if a method performs better in the success rate of binding mode prediction,it also performs better in the average RMSD of ‘hits’.That means,the performance comes from both the number and the quality of successful predictions.
Table 1 The success rates (%) predicted by our CHDOCK and three other symmetric docking programs on our protein docking benchmark of 212 Cnsymmetric complexes when the top 1,10,and 100 predictions were consideredBound docking Unbound docking Method Top 1 Top 10 Top 100 Top 1 Top 10 Top 100 CHDOCK 55.19 72.17 90.57 30.66 44.81 68.40 M-ZDOCK 45.76 65.09 89.15 26.42 36.79 66.51 SAM 35.85 54.25 84.91 19.34 31.60 63.68 SymmDock 16.04 31.60 67.45 11.79 30.66 58.49
Fig.1 The success rates of our CHDOCK and three other symmetric docking methods in binding mode predictions on our protein docking benchmark of 212 Cnsymmetric complexes for bound docking(A)and unbound docking(B).For each method,from left to right are for the results of top 100,10,and 1 prediction,respectively
Table 2 The average LRMSD(?)of ‘hit(s)’predicted by our CHDOCK and three other symmetric docking programs on our protein docking benchmark of 212 Cnsymmetric complexes when the top 1,10,and 100 predictions were consideredBound docking Unbound docking Method Top 1 Top 10 Top 100 Top 1 Top 10 Top 100 CHDOCK 1.10 1.51 2.11 2.54 3.12 4.07 M-ZDOCK 2.27 2.80 3.46 3.26 3.86 5.07 SAM 2.28 2.88 4.10 3.43 4.03 5.48 SymmDock 3.40 4.42 5.72 5.15 5.61 6.43
Performance of scoring function
Fig.2 The average RMSD of first ‘hit(s)’ of our CHDOCK and three other symmetric docking methods tested on our protein docking benchmark of 212 Cnsymmetric complexes for bound docking(A)and unbound docking(B).For each method,from left to right are for the results of top 100,10,and 1 prediction,respectively
To investigate the performance of our scoring function,we also tested our pure FFT-based docking,named CHDOCK_lite,on the benchmark,which only uses the shape complementarity to filter and sort docking docking results for bound docking and unbound docking are shown in Fig.3.It can be seen from the figure that CHDOCK performed much better than CHDOCK_ the help of our scoring function ITScorePP (Huang and Zou 2008),the success rate of bound docking for top 1 prediction increased from 21.70% to 55.19% and for unbound docking,the success rate increased from 11.32% to 30.66%.The great improvement of CHDOCK compared to CHDOCK_lite demonstrates the important role of our scoring function.
Discussions
CHDOCK and M-ZDOCK are both the three-dimensional(3D) FFT-based docking algorithms and adopt the similar sampling ,the difference between CHDOCK and M-ZDOCK is that CHDOCK adopts a better shape complementarity score LSC (Yan and Huang 2018) and a more powerful scoring function ITScorePP(Huang and Zou 2008).In our previous study on hetero protein complexes (Yan and Huang 2018),LSC has shown its better performance than PSC(Chen and Weng 2003) used in also showed a better performance in scoring decoys and finding the near native structures (Huang and Zou 2008).Therefore,the better performance of CHDOCK than M-ZDOCK would be attributed to both the shape complementarity score LSC and our scoring function CHDOCK has achieved better performance than the other three docking programs,the success rate for top 1 prediction is still not high,especially for unbound are much room to improve the existing methods and develop new docking programs in the future.
文章来源:《采矿与安全工程学报》 网址: http://www.ckyaqgcxb.cn/qikandaodu/2020/0925/363.html
上一篇:CNKI推出《中国高被引图书年报》
下一篇:北京经济学院劳动保护工程系改名