2.1. Dynamics Equilibrium and Structural Fluctuation
To check structural stability of BACE1 during three separate MD simulations, root-mean-square deviations (RMSDs) of backbone atoms in BACE1 relative to the initially optimized structures were calculated based on three separate MD trajectories and the results were provided in supporting information
Figure S1. It is noted that all four systems have reached the equilibrium after 300 ns of three separate MD simulations. The greater frequency distributions of the
apo, 60W-, 954- and 60X-bound BACE1 are respectively located at 2.11, 2.29, 1.93 and 2.06 Å, moreover the distribution shape of the 954- and 60X-bound BACE1 moves toward the left relative to that of the
apo BACE1 while the distribution shape of the 60W-bound BACE1 moves toward the right (
Figure 2A). Thus, binding of 954 and 60X weakens the structural fluctuation of BACE1 compared to the
apo state while the presence of 60W increases the structural fluctuation of BACE1. The RMSDs of heavy atoms from three inhibitors relative to the initially optimized structures were also computed to check their stability in binding pocket of BACE1 and the frequency distributions were depicted at
Figure 2B. The RMSD of 60W is distributed at two peak values of 0.68 and 1.77 Å while the RMSDs of 954 and 60X are respectively populated at the peak values of 3.05 and 1.04 Å (
Figure 2B). These results indicate that 60W and 60X have lower structure fluctuations in binding pocket of BACE1 than 954, implying that 60W and 60X have higher structure stability in binding pocket of BACE1 than 954.
To examine inhibitor’s binding-mediated impacts on structural flexibility of BACE1, root-mean-fluctuations (RMSFs) were computed by using the coordinates of the atoms Cα kept at the SMT (
Figure 2C). It is found that the
apo and bound states of BACE1 share similar flexible and rigid regions. The RMSF difference between the
apo and bound states was also calculated by utilizing the equation
, in which
,
and
represent the RMSF difference, the RMSFs of the bound and
apo states (
Figure 2D). The presence of the three inhibitors weakens the structural flexibility of the regions D1 (residues 110-123) and D4 (residues 178-202), which makes these two regions be more rigid than the
apo BACE1 (
Figure S2). However, binding of the three inhibitors strengthens the structural flexibility of the regions D2 (residues 146-153) and D3 (residues 164-178) relative to the
apo BACE1 (
Figure 2D and
Figure S2). The presence of 60W and 60X totally enhances the structural flexibility of the region D5 (residues 217-245) but binding of 954 weakens that of this region relative to the
apo BACE1 (
Figure 2D). By comparison with the
apo BACE1, binding of 60W increase the structural flexibility of the region D6 (residues 363-387) while the presence of 954 and 60X slightly weakens that of this region (
Figure 2D and
Figure S2)
To access stability of secondary structures for BACE1, the combination of the program CPPTRAJ and the DSSP second structure analysis [
64] were used to probe the stability of the second structure of BACE1 from the
apo and bound states in three separate MD simulations. The time evolutions of the secondary structures for the
apo, 60W-, 954- and 60X-bound BACE1 were displayed in
Figure 3A–C, S3, S4 and S5, individually. It is observed that the secondary structure of the
apo BACE1 is stable through three separate MD simulations (
Figure 3A,B). By comparison with the
apo BACE1, the secondary structures of the 60W-, 954- and 60X-bound BACE1 do not generate obvious changes (
Figures S3–S5), indicating that binding of inhibitors hardly affects the stability of the secondary structures of BACE1. To understand influences of inhibitor binding on the structure compact extents of BACE1, the gyrations of BACE1 in the
apo and bound states were estimated based on the SMT and their frequency distributions were provided at
Figure 3D. The gyration of the
apo BACE1 is distributed at two peaks of 20.87 and 21.02 while the ones of the 60W-, 954- and 60X-associated BACE1 are separately populated at the single peaks of 21.17, 20.99 and 20.99 Å, furthermore the distribution shapes of gyrations for BACE1 in three bound states totally move toward the right by referencing to the
apo BACE1 (
Figure 3D), implying that the presence of the three inhibitors slightly leads to a more incompact structures of BACE1.
Based on the current analyses, binding of the three inhibitors affects structural fluctuations and compact extents of BACE1 but hardly changes the stability of secondary structures for BACE1. The difference in structures of 60W, 954 and 60X also leads to their different stability in binding pocket of BACE1. Meanwhile, binding of inhibitors generates obvious effect on structural flexibility of BACE1, which also implies possible hot spots of inhibitor-BACE1 interactions. These current results basically agree with that of a previous work [
16].
2.2. Conformational Changes of BACE1 and Free Energy Profiles
To clarify inhibitor-mediated changes in correlated motions of BACE1, DCCMs were computed by means of the coordinates of the atoms Cα in BACE1 and the results were exhibited at
Figure 4. The color bar coded by different colors was employed to embody the contents of the correlated motions between residues of BACE1. For the
apo BACE1, several strongly correlated motions are observed: (1) the region R1 describes strong PCMs of the N-terminal from BACE1 relative to itself (
Figure 4A), (2) the region R2 reflects strong ACMs of residues 206-265 relative to residues 115-158 and the region R3 characterizes strong ACMs between residues 278-331 and 146-200, (3) the region R4 embodies strong ACMs of residues 347-418 relative to the N-terminal of BACE1 and (4) the region R5 describes strong PCMs of residues 266-328 relative to themselves (
Figure 4A). Compared to the
apo BACE1, binding of the three inhibitors highly weakens the PCMs occurring at the regions R1 and R5 (
Figure 4B–D). Meanwhile, binding of 60W, 954 and 60X also reduces the ACMs of the regions R2 and R3 by referencing to the
apo BACE1 (
Figure 4B–D). Differently, binding of 60W and 954 obviously abates the ACMs of the region R4 by comparison with the
apo BACE1 (
Figure 4B,C) but the presence of 60X slightly strengthens the ACMs of this region (
Figure 4D). The aforementioned regions corresponding to evident alterations of correlated motions are possibly involved in hot interaction spots of inhibitors with BACE1.
To reveal impacts of inhibitor binding on concerted movements of structural domains in BACE1, PCA was carried out by using the CPPTRAJ in Amber 20. The first eigenvector from the PCA was visualized by means of the software VMD [
65] and the results were depicted
Figure 5. It is observed that inhibitor binding yields evident influences on the collective motions of the α helix α1 and the loop L2 (
Figure 5). In the
apo BACE1, the α1 and L2 generate a parallel concerted motion with the same direction (
Figure 5A). Compared to the
apo BACE1, binding of 60W enhances the concerted motion of α1 and L2 and obviously alters the direction of the concerted motion for the L2 (
Figure 5B). By comparison with the
apo BACE1, the presence of 954 not only leads to a completely opposite motion direction of the α1 and L2 but also inhibits the motion amplitude of the L2, which produces a tendency of the L2 away from the α1 (
Figure 5C). By referencing to the
apo BACE1, binding of 60X not only changes the concerted motion direction of the α1 and L2 but also apparently weakens the concerted motion of the L2 (
Figure 5D). Binding of 60W and 954 slightly inhibits the concerted movement of the loop L1 relative to the
apo BACE1 but the presence of 60X slightly strengthens the concerted motion of this loop (
Figure 5B–D). In addition, binding of the three inhibitors weakens concerted motion of the loop L3 compared to the
apo BACE1 (
Figure 5B–D).
To unveil free energy profiles of the BACE1 conformation changes caused by inhibitor binding, FELs were created by using the projections (PC1 and PC2) of the SMT on the first two eigenvectors as reaction coordinates (RCs) and the presentative structures relating with free energy profiles were depicted in
Figure 6 and S6. The projections of MD trajectories can rationally reflects conformational changes of BACE1, which is the cause for our selecting them as RCs.
In the case of the
apo BACE1, three separate MD simulations capture three free energy valleys (EVs), including EV1, EV2 and EV3 (
Figure S6A). According to the color bar, three EVs are located at the valley bottoms of the same depth (
Figure S6A). Three presentative structures of the
apo BACE1 in the EV1-EV3 were superimposed together (
Figure S6B). The results suggest that the domains D4-D6 yield obvious deviations from each other. The structure domains D1 and D3 generate slight deviations and the D2 produces the slight sliding (
Figure 6B)
Compared to the
apo BACE1, binding of 60W and 954 only results in two EVs (
Figure 6A,D), which is less than the energy states of the
apo BACE1. This result implies that binding of 60W and 954 induces conformational arrangement of BACE1 relative to the
apo BACE1. The representative structures of the 60W- and 954-bound BACE1 situated at the EV1 and EV2 were superimposed together to probe structural difference (
Figure 6B,E). By comparison with the
apo BACE1, binding of 60W and 954 reduces the structural deviations of the structural domains D1, D3, D4 and D5 (
Figure 6B,E and S6B). Although binding of 60W weakens the structural deviation of the D2 and D6 relative to the
apo BACE1, the association of 954 leads to great deviation of the D2 and D6 (
Figure 6B,E and S6B). As shown in structural alignment of inhibitors 60W and 954 falling into the EV1 and EV2 (
Figure 6C,F), 60W and 954 produce slightly sliding between two energy states, which possibly affects binding of these two inhibitors to BACE1. As for the 60X-bound BACE1, three EVs are detected through the entire MD simulations (
Figure 6G). Although binding of 60X does not alter the number of the EVs relative to the
apo BACE1, the presence of 60X enhances the energy barrier between the EV3 and two states EV1 and EV2 by referencing to the apo BACE1 (
Figure 6G and S6A), which correspondingly increases the difficulty of the transitions between the EV3 and the EV1 and EV2. According to structural superimposition of the 60X-bound BACE1 trapped at the EV1-EV3 (
Figure 6H), except for the structural domains D1, D3, D4 and D5, binding of 60X evidently increases the deviation of the D2 and D6 among three EVs compared to the apo BACE1. The structural alignment of 60X falling into the EV1-EV3 indicates that 60X yields slight deviations among three energetic states (
Figure 6I).
Based on the aforementioned calculations of DCCMs, PCA and analyses of FELs, binding of inhibitors changes correlated motions between residues, affects concerted movements of structural domains and alters free energy profiles of BACE1. Some of the structural domains affected by inhibitor binding are located near the binding pocket, hence conformational changes caused by binding of inhibitors in turn alter the activity of BACE1. Several previous works also revealed similar results [
20,
61], which is in basic agreement with our current work.
3.3. Comparative Calculations of Binding Free Energies
To access binding ability of 60W, 954 and 60X to BACE1, the SIE method was applied to calculate binding affinities to the three inhibitors to BACE1 by using 500 snapshots extracted from the equilibrated section of three separated MD simulations, namely for the SMT, in a time interval of 1.8 ns. The calculated results were listed in
Table 1. It is observed that the rank of binding affinities predicted by the SIE method is in consistence with that of the experimental values, which indicates that our current free energy analyses are rational and reliable.
According to
Table 2, the components of binding affinities predicted by the SIE method mainly consist of the intermolecular Coulomb interactions (
), the van der Waals ones (
), the reaction energy (
) and the energy changes in the molecular surface area upon binding(
). The energy contributions favoring binding of inhibitors are those from the van der Waals interactions between binding partners (-42.50 to -54.53 kcal·mol
-1), the intermolecular Coulomb interactions (-12.55 to -16.18 kcal·mol
-1) and the energy contributions relating with the changes in the molecular surface (-8.05 to -10.63 kcal·mol
-1). The reaction energies fluctuate at a range of 21.27 to 24.77 kcal·mol-1 and this component provide an unfavorable force for the inhibitor bindings, which is also revealed by the previous work [
49,
66,
67]. On the basis of
Table 2, the unfavorable reaction energies of three inhibitor-BACE1 complexes are partially compensated by the favorable intermolecular Coulomb interaction. Meanwhile, the intermolecular van der Waals interactions also contribute partial compensation to this unfavorable effect. Among three inhibitors, 60W shows the strongest binding ability to BACE1 (-8.82 kcal·mol
-1) while 954 has the weakest binding ability to BACE1 (-7.30 kcal·mol
-1), which suggests that small structure difference among the three inhibitors impacts their binding ability to BACE1.
To comparatively study binding strength of 60W, 954 and 60X to BACE1, MM-GBSA method was adopted to predict binding free energies of three inhibitor-BACE1 complexes based on 500 snapshots extracted from the equilibrated section of three separated MD simulations, namely for the SMT, in a time interval of 1.8 ns. Because of expensive time in the entropy calculation, 100 snapshots taken from the above mentioned 500 snapshots were employed to perform the calculation of the entropy contributions to inhibitor-BACE1 binding. The MM-GBSA calculations are possibly involved in multiple generalized Born (GB) models. To understand influences of different GB models on the predicted results, four GB models, indicated by IGB=1, IGB=2, IGB=5 and IGB=66, were chosen to estimate binding free energies of the three inhibitors to BACE1. The empirical parameters involved in calculations of four GB models were given in
Table 2, which includes two empirical parametersγand β together with the radii types. Binding free energies and their components computed by the MM-GBSA method were listed in
Table 3.
Binding free energies are mainly composed of five components, including the van der Waals interactions (
), the electrostatic interactions (
), the polar solvation free energy (
), the non-polar solvation free energy (
) and the entropy contributions (
), which is shown in
Table 3. From free energy components,
,
and
are favorable for inhibitor-BACE1 binding but
and
impair inhibitor-BACE1 associations (
Table 3). The hydrophobic interactions (
) formed by the sum of
and
are favorable for inhibitor-BACE1 binding. The polar interactions (
) formed by the sum of
and
provide different-type force for inhibitor-BACE1 association. In details,
predicted by the models of IGB=1, IGB=2 and IGB=66 is unfavorable for the inhibitor-BACE1 binding while that predicted by the model of IGB=5 contributes favorable forces to the inhibitor-BACE1 associations (
Table 3). The sum of four components
,
, and
constructs the enthalpy contributions (∆H) to the inhibitor-BACE1 binding. Based on
Table 3, the GB models used for calculations of MM-GBSA only produce evident impacts on polar solvation free energies and the selection of the empirical parameters γ and β obviously affects calculations of non-polar solvation free energies. By comparison on four GB models, the GB model of IGB=5 leads to the weakest polar solvation free energies for all inhibitors but that of IGB=66 yields the strongest polar solvation free energies (
Table 3). Correspondingly, the GB model of IGB=5 generates the strongest enthalpy contributions to inhibitor-BACE1 association but that of IGB=66 produces the weakest enthalpy contributions to inhibitor-BACE1 binding. As a result, the selection of the GB models brings a vital impact on predictions of inhibitor-BACE1 binding free energies.
For our used GB models, binding free energies of 60W, 954 and 60X to BACE1 estimated with the GB model of IGB=2 is mostly close to the experimental values. Differently, binding free energies of the three inhibitors to BACE1 calculated through the GB models of IGB=1, 5 and 66 highly deviate from the experimental results. Meanwhile, the rank for binding free energies of 60W, 954 and 60X in four GB models are also in good agreement with that for the experimental values, verifying that our current results are reliable and rational. Based on the aforementioned analyses, the results calculated by the GB model of IGB=2 are utilized to access binding difference of the three inhibitors to BACE1. The electrostatic interactions of 60W and 60X with BACE1 are respectively strengthened by 7.98 and 5.63 kcal/mol relative to that of 954 with BACE1 but unfavorable polar solvation free energies of the 60W- and 60X-BACE1 complexes are raised by 11.42 and 3.99 kcal/mol by comparison with that of the 954-bace1 complex. On the whole, the polar interaction of 60W with BACE1 is increased by 3.44 kcal/mol by referencing to that of 954 with BACE1 while the polar interaction of 60X with BACE1 is reduced by 1.64 kcal/mol. The hydrophobic interaction of 60W with BACE1 strengthened by 13.3 kcal/mol compared to that of 954 with BACE1, but the hydrophobic interaction of 60X with BACE1 hardly changes relative to that of 954 with BACE1. As a result, the enthalpy contributions to the 60W- and 60X-BACE1 binding is improved by 9.86 and 1.63 kcal/mol by referencing to that of the 954-BACE1 binding. In addition, the unfavorable entropy contributions to the 60W- and 60X-BACE1 binding is increased by 3.92 and 0.34 kcal/mol relative to the 954-BACE1 binding. In summary, the binding ability of 60W and 60X to BACE1 is strengthened by 5.94 and 1.29 kcal/mol compared to that of 954 to BACE1 (
Table 3). Therefore, although structural difference of the three inhibitors is tiny, their binding ability to BACE1 produces the bigger difference according to our current calculations, which should owe to the conformational changes caused by their binding.
By combination of the SIE and MM-GBSA calculations, it is found that hydrophobic interactions provide a key contribution to inhibitor-BACE1 binding, which is in good agreement with the previous reports [
16]. Thus, the rational optimization on inhibitor-BACE1 hydrophobic interactions is of high significance for successful design of clinically available inhibitors toward BACE1. Based on this issue, more attentions should be paid to hydrophobic interactions of inhibitors with BACE1.
2.4. Analyses of Inhibitor-BACE1 Interaction Networks
To obtain atomic-level insights into interaction modes of inhibitors with BACE1, residue-based free energy decomposition method was applied to estimate inhibitor-residue interaction spectrum of three inhibitor-BACE1 complexes (
Figure 7). The contributions from the sidechains and backbones of residues to inhibitor-BACE1 associations were provided in
Table 4. The hydrogen bonding interactions (HBIs) between inhibitors and residues of BACE1 were analyzed by using the program CPPTRAJ and the results were listed in
Table 5. The geometric information with regard to inhibitor-residue interactions was depicted in
Figure 8. Meanwhile, the frequency distributions of the distances relating with inhibitor-residue interactions were also calculated and the results were displayed in
Figure 9.
For the 60W-BACE1 complex, 60W produces the interactions stronger than -1.0 kca/mol with six residues of BACE1, including L91, D93, S96, V130, Q134 and I179 (
Figure 7A and 7D). The three residues D93, S96 and V130 are situated near the hydrophobic rings R1 and R2 of 60W (
Figure 8A). Hence, D93 forms the CH-O interactions with these two rings, S96 generates the CH-π and CH-O interactions with the ring R1 and V130 yields the CH-π interaction with the ring R1 of 60W (
Figure 8A). According to
Table 4, the energetic contributions of S96 and V130 to the 60W-BACE1 binding mostly arise from the sidechains of these two residues. Additionally, the carbonyl of D93 generates four HBIs with the ring R2 of 60W and their occupancy is higher than 46.7% (
Table 5 and
Figure 8B), meanwhile the favorable 60W-D93 interaction mainly comes from the electrostatic interaction of the sidechain of D93 (
Table 4). On the whole, D93, S96 and V130 provide energy contributions of -2.04, -1.13 and -1.75 kca/mol to the 60W-BACE1 binding, respectively (
Figure 7A,D and
Table 4). The distances for the mass centers of the sidechains of V130 and S96 away from that of the ring R1 are respectively distributed at 4.03 and 4.03 Å (
Figure 9A), which verifies the interactions of these two residues with 60W. The distance between the mass center of the carbonyl of D93 and that of the ring R2 in 60W is populated at 6.09 Å, which agrees with the weak CH-O interaction of D93 with 60W (
Figure 9A). The residues Q134 and I179 are next to the ring R3 of 60W and these two residues form the CH-π interactions with the ring R3 of 60W (
Figure 7A,D and
Figure 8A). As shown in
Table 4, the van der Waals interactions of the sidechains from Q134 and I179 with the ring R3 of 60W contributes the most forces to the 60W-BACE1 association. The distance of the carbon atom from Q134 and the mass center of the alkyl group in I179 away from mass center of the ring R3 in 60W are situated at 4.03 and 3.66 Å, respectively (
Figure 9A). As a result, the two residues Q134 and I179 separately provide the interaction energies of -2.58 and -2.26 kcal/mol for binding of 60W to BACE1 (
Figure 7A,D). The interaction energy of L91 with 60W is -1.72 kcal/mol (
Figure 7A,D), which structurally stems from the CH-π interaction between the alkyl group of L91 and the ring R4 of 60W (
Figure 8A). More interestingly, the energy contribution of L91 is mainly provided by the van der Waals interactions between the sidechain of L91 and the ring R4 of 60W (
Table 4). The distance between the mass center of the alkyl group from L91 and that of the ring R4 in 60W is distributed at 3.84 Å (
Figure 9A), which demonstrates the existence of the CH-π interaction between 60W and L91.
With respect to the 954-BACE1 compound, five residues are involved in interactions stronger than -1.0 kcal/mol with the inhibitor 954 and these residues include L91, Y132, W137, F169 and I179 (
Figure 7B,D). The interaction energies of Y132 and W137 with 954 are -1.88 and -1.04 kcal/mol, individually, which structurally agrees with the π-π interactions of the phenyl group in Y132 with the ring R2 of 954 and the hydrophobic ring of W137 with the ring R1 of 954 (
Figure 8C). The distances of the mass centers for the hydrophobic rings of Y132 and W137 away from that of the rings R2 and R1 from 954 are located at 4.87 and 5.09 Å (
Figure 9B), which further supports the interactions of these two residues with 954. Based on
Table 4, the energy contributions of Y132 and W137 to the 954-BACE1 binding are mostly provided by the van der Waals interactions of the sidechains from Y132 and W137 with 954. The hydrophobic groups L91, F169 and I179 are located near the ring R4 of 954 (
Figure 8C). Therefore the alkyl group of L91, the phenyl group of F169 and the alkyl group of I179 tend to generate the CH-π, π-π and CH-π interactions with the R4 of 954. The distances of the mass centers of the sidechains in L91, F169 and I179 away from that of the ring R4 in 954 are respectively populated at 4.86, 6.42 and 4.24 Å (
Figure 9B), which verifies the hydrophobic interactions of these three residues with 954. On the whole, L91, F169 and I179 contribute the interaction energies of -1.14, -1.44 and -1.91 kcal/mol to the 954-BACE1 binding (
Figure 7B,D and
Table 4). More importantly, the interaction energies of L91, F169 and I179 with 954 mostly origin from the van der Waals interactions of the sidechains in these three residues with 954 (
Table 4). In addition, the carbonyl group of D93 forms four HBIs with the ring R3 of 954 and the occupancy of these four hydrogen bonds are higher than 46.9% (
Table 5 and
Figure 8D). However, D93 only provides an energy contribution of -0.77 kcal/mol (
Figure 7B,D), which mainly stems from the electrostatic interaction between the sidechain of D93 and 954 (
Table 4).
With regard to the 60X-BACE1 complex, 60X yields the interactions stronger than -1.0 kcal/mol with five residues D93, V130, Y132, W137 and I179 in BACE1 (
Figure 7C,D). The hydrophobic groups of I179 and Y132 are adjacent to the ring R2 of 60X (
Figure 8E), hence the alkyl group of I179 forms the CH-π interactions with the ring R2 of 60X and the phenyl group of Y132 generates the π-π interaction with the ring R2 of 60X (
Figure 8E). The distances of the mass centers for the hydrophobic groups of Y132 and I179 away from that of the ring R2 in 60X are individually populated at 4.52 and 5.13 Å (
Figure 9C), which further supports the interactions of Y132 and I179 with 60X. Y132 and I179 separately contribute the interaction energies of -2.0 and -2.31 kcal/mol to the 60X-BACE1 binding (
Figure 7C,D), furthermore they mostly come from the van der Waals interactions of the sidechains in Y132 and I179 with the ring R2 of 60X (
Table 4). Two residues V130 and W137 produce the interactions of -1.28 and -1.29 kcal/mol with 60X (
Figure 7C,D), which is in good agreement with the CH-π interaction of the alkyl group from V130 and the π-π interaction of the hydrophobic ring of W137 with the ring R3 of 60X (
Figure 8E). The distances of the mass centers for the sidechains of V130 and W137 away from that of the ring R3 in 60X are separately distributed at 4.32 and 6.13 Å (
Figure 9C), implying the existence of the interactions of V130 and W137 with 60X. More importantly, the energy contributions of V130 and W137 to the 60X-BACE1 association are mainly provided by the van der Waals interactions of the sidechains of V130 and W137 with 60X (
Table 4). Besides, the carbonyl group of the residue D93 not only produces the CH-O interactions with the ring R1 of 60X but also forms four HBIs with the occupancy higher than 48.7% with the ring R2 of 60X (
Figure 8F and
Table 5). The distance between the mass center of the carbonyl group of D93 and that of the ring R2 in 60X is distributed at 4.72 Å (
Figure 9C), implying the existence of the CH-O interactions of D93 with 60X.
Based on the aforementioned description, three inhibitors form hydrophobic interactions with L91, S96, V130, Q134, W137, F169 and I179 and the energy contributions of these residues to inhibitor binding mostly come from the interactions of their sidechains with inhibitors. The residue D93 produces four HBIs with inhibitors and these HBIs are formed between the carbonyl group (the sidechain) of D93 and inhibitors. It is concluded that the sidechains of the above mentioned residues play key roles in binding of inhibitors to BACE1. Therefore, it is of high significance to rationally optimize the interactions of inhibitors with the sidechains of key residues in BACE1 for design of efficient inhibitors toward BACE1.