Exploring the Binding Mechanism between Human Profilin (PFN1) and Polyproline-10 through Binding Mode Screening

Leili Zhang; David R. Bell; Binquan Luan; Ruhong Zhou

The rupture works calculated from two rupture speeds (1nm/ns and 0.1 nm/ns) both correlate well with the PMF differences (Figure 4 A, Figure 4 C), where R 2 is 0.49 for 1 nm/ns SMD, and 0.67 for 0.1 nm/ns SMD. The absolute values of rupture works from 0.1nm/ns SMD (<-15 kcal/mol) are closer to the PMF differences (see Table S1 for details), compared to the range (-10 ~ -30 kcal/mol) from 1 nm/ns SMD. This is in agreement with previous practice on free energy estimations from SMD simulations. 47 For our purpose, instead of following the common sense of "the slower the better" in the field (0.01 or even 0.001 nm/ns), we use a relatively fast rupture speed (1 or 0.1 nm/ns) to rank the docked structures. For example, each simulation only takes ~2.5 ns or ~25 ns for SMD with 1nm/ns or 0.1nm/ns rupture speed, respectively. This means that from 2.5 ns×10 replicas = 25 ns of simulations in total, a trend in PMFs can be predicted by SMD with high confidence, as compared to the 50 ns of MD simulation sampling a local well on the free energy landscape, or 20×20 = 400 ns of umbrella sampling for numerous binding modes from one PPI pair. Moreover, the replica numbers can be reduced to 6 to reach a comparable R 2 (see Figure S8 A, Figure S8 B) which further decreases the total simulation time to 15 ns (for 1 nm/ns SMD).