Figure 1.
Identifying surface binding pockets.
(A) Bcl-XL (grey surface) is shown in complex with an inhibitor (blue sticks). The protein surface features a large pocket (red spheres) complementary in shape to the inhibitor. (B) Deep pocket volumes of surface pockets at protein interaction sites harboring a bound inhibitor (red line) are larger than those found elsewhere on the protein surface (black line). Data are collected from a test set of seven proteins, each of which has been solved in complex with a small-molecule inhibitor (Bcl-XL, IL-2, FKBP12, HPV E2, ZipA, MDM2, and the BIR3 domain of XIAP).
Table 1.
Our test set is comprised of proteins for which structures have been solved both alone and in complex with a small-molecule inhibitor bound to the protein interaction site.
Figure 2.
Surface pockets emerge only at druggable sites.
Volumes of surface pockets are shown from conformations generated with no biasing potential (left) and upon inclusion of a “pocket opening” biasing potential (right) for each of the seven proteins that comprise our test set. Surface pockets occur at druggable protein interaction sites (solid red lines) more frequently than elsewhere on the protein surface (dashed black lines). (A) Bcl-XL. (B) IL-2. (C) FKBP12. (D) HPV E2. (E) ZipA. (F) MDM2. (G) BIR3 domain of XIAP.
Figure 3.
Energetic analysis of Bcl-XL pocket opening.
(A) Conformations generated without the use of a biasing potential (solid green line) show a similar distribution of energies to those generated with the biasing potential at a randomly selected target residue (dashed black line); increasing the strength of the biasing potential here leads to conformations with higher energies (solid black line). In contrast, application of the biasing potential at the protein interaction site (red lines) leads to conformations with a distribution of energies that strongly overlaps with those energies of conformations sampled in the unbiased simulations, suggesting that these conformations represent low-energy states accessible to the unbound protein. (B) A scatterplot showing the deep pocket volume for conformations generated with the biasing potential applied to one of the random sites (moderate bias in cyan, strong bias in blue) or to the protein interaction site (moderate bias in red, strong bias in orange). Low-energy conformations containing large pockets are sampled only if the biasing potential it is applied at the protein interaction site; while large pockets are sampled using the strong bias at random sites, these conformations have considerably higher energy. All energies shown here were evaluated in the absence of the biasing potential, for fair comparison.
Figure 4.
Representative conformations of Bcl-XL
. An unbound crystal structure (pink), an inhibitor-bound crystal structure (green, with inhibitor shown in sticks), and a low-energy conformation generated from the unbound crystal structure using the biasing potential (cyan, with target residue in red) are shown. (A) The overall protein architecture is preserved amongst all three; movement of the helix in the foreground upon binding is not recapitulated in the pocket-opened conformation. (B–D) The pocket revealed in this low-energy conformation nonetheless strongly resembles the surface pocket in the bound crystal structure, and even bears shape-complementarity to the inhibitor. The identity of the inhibitor was not used in generating this conformation, but was added retrospectively for visual comparison.
Figure 5.
Opening pockets shifts the conformational ensemble towards inhibitor-bound structures.
Distributions of RMSD over interface atoms (iRMSD) to the closest inhibitor-bound crystal structure for conformations generated with (red lines) or without (black lines) the biasing potential. The iRMSDs from the unbound structure to the most and least similar inhibitor-bound crystal structures are also indicated (dashed brown vertical lines). For all seven proteins comprising the test set, the biased simulations produced conformations closer to an inhibitor-bound crystal structure than the unbiased simulations.
Figure 6.
Distinguishing druggable from functional sites on survivin.
(A) The crystal structure of survivin, showing the protein interaction site (red) and the distal druggable site identified by NMR (blue) [14]. (B) Volumes of surface pockets are compared for conformations generated with the biasing potential applied at random surface residues (dashed black lines), applied at the protein interaction site (dashed red lines), and applied at the distal druggable site (solid blue lines). Pockets emerge at the druggable site but not elsewhere on the protein surface.