注册 English

Discovery of 4-Thiazol-N-(pyridin-2-yl)pyrimidin-2-amine as Novel Cyclin-dependent Kinases 4 and 6 Dual Inhibitors via 3D-QSAR and Molecular Simulation

Le FU ,  Li-Nan ZHAO ,  Hong-Mei GUO ,  Na YU ,  Wen-Xuan QUAN ,  Yi CHEN ,  Mao SHU ,  Rui WANG ,  Zhi-Hua LIN

downloadPDF
Citation:  Le FU, Li-Nan ZHAO, Hong-Mei GUO, Na YU, Wen-Xuan QUAN, Yi CHEN, Mao SHU, Rui WANG, Zhi-Hua LIN. Discovery of 4-Thiazol-N-(pyridin-2-yl)pyrimidin-2-amine as Novel Cyclin-dependent Kinases 4 and 6 Dual Inhibitors via 3D-QSAR and Molecular Simulation[J]. Chinese Journal of Structural Chemistry, 2022, 41(3): 220310. doi: 10.14102/j.cnki.0254-5861.2011-3270 shu

Discovery of 4-Thiazol-N-(pyridin-2-yl)pyrimidin-2-amine as Novel Cyclin-dependent Kinases 4 and 6 Dual Inhibitors via 3D-QSAR and Molecular Simulation

English

  • 1. INTRODUCTION

    During the ordered cell cycle, cell undergoes a transition from quiescence or cytoplasmic division to cell proliferation, and checkpoints then further screen for stable genomes. It consists of four consecutive phases: G1 (pre-DNA synthesis), S (DNA synthesis), G2 (predivision) and M (cell division)[1]. Cells remain quiescent by deciding to enter S phase or signaling pathways that link extracellular signals to G1 phase. At the same time, the G1 phase depends on retinal tumor cells to inhibit the expression of protein (Rbs). Rbs protein works by inhibiting the E2 promoter binding factor (E2F) transcription factor, which is essential for G1 to S phase transition. Released E2F transcription factor phosphorylates CDKs retinoblastoma protein (Rb) that promotes the transcription of genes encoding proteins (Cyclin A and Cyclin E) required for DNA replication in G1 phase. CDK4 and CDK6 are the main kinases of G1 phase phosphatization in mammalian cells[2].

    Previous researches have demonstrated the CDK4/6-Rb-E2F way is blocked in ninety percent of cancers[3]. In addition, genetic researches have demonstrated that inactivation of CDK4/6 merely applies the proliferation of particular cell and that CDK4/6 is optional for mitotic cycle. Therefore, CDK4/6 is potential targets for the treatment of cancer[4]. As approved by the FDA first CDK4/6 inhibitors, palbociclib has approved for metastatic breast cancer combining with letrozole or fluvistat[5, 6]. This stands for a significant scientific advance in the field. However, the discovery of new skeleton CDK4/6 inhibitors and their derivatives remains a challenge.

    Since major inhibitors combine to conserved sites of ATP-binding, target selectivity is still not conquered. For example, many clinically used CDK4/6 inhibitors (palbociclib, abemaciclib and ribociclib shown in Fig. 1) disclose some additional targets for a near kinome-wide scope[7]. The off-targets can compromise drug safety and cause differential cellular efficacy and clinical responses. For instance, ribociclib and palbociclib are primarily myelotoxic. Instead, abemaciclib leads to plentiful reactions. Palbociclib and ribociclib should be used periodically, while abemaciclib should continue to be used because of the relatively less proof of neutrophil reduction[8]. Jointly, these evidences suggest that the introduction of new pharmaceutical groups may increase kinase selectivity, thereby reducing adverse side effects.

    Figure 1

    Figure 1.  Structures of CDK4/6 inhibitors
    DownLoad: Full-Size Img PowerPoint

    In our ongoing research for novel and highly CDK-targeting drugs, we used computational ways including 3D-QSAR, drug design, molecular docking and dynamic simulation to study 4-thiazol-N-(pyridin-2-yl)pyrimidin-2-amine CDK4/6 inhibitors with a view to designing novel, highly selective and orally bioavailable dual CDK4/6 inhibitors.

    2. COMPUTATIONAL METHODS

    2.1 Data source and processing

    The structures of dual CDK4/6 inhibitors were got from the literatures[9-11]. In general, the lowest-energy conformation is more stable and is theoretically thought to be active in small molecules. All molecules were optimized by adding Gasteiger-Huckel charge, Powll energy gradient mean and Tripos force field in SYBYL-X2.0 software[12]. All parameters were default except the maximum optimization limit and convergence criterion. They were set as 10000 times and 0.005 kcal/mol, respectively[13]. To build CoMFA/CoMSIA model, the dataset was randomly divided into training set (quantity = 40) and test set marked with "*" in Table 1S. At the same time, Table 1 contains active data (pKi = –logKi)[14].

    Table 1

    Table 1.  Chemical Structures of Dual CDK4/6 Inhibitors
    DownLoad: CSV

    2.2 Molecular alignment

    Molecular alignment is a necessary preparation for obtaining reliable 3D-QSAR model[15]. Compound database alignment is the process of collecting molecular descriptors in pharmaceutical molecules and simulating action points[16]. We selected compound 16 as the template, which has the highest activity. After selecting violet-red common Skeleton (Fig. 2a)[17], the superimposed structures of aligned compounds are shown in Fig. 2b.

    Figure 2

    Figure 2.  (a) Alignment of dual CDK4/6 inhibitors; (b) Structure of compound 16
    DownLoad: Full-Size Img PowerPoint

    2.3 3D-QSAR model modeling

    Using CoMFA and CoMSIA methods generated 3D-QSAR models. Using Coulomb method to compute the electrostatic field contribution and using Lennard-Jones method to compute the stereo field contribution in CoMFA model[18]. The CoMSIA model is established by introducing additional hydrogen bond application field, hydrogen bonding body field and hydrophobic field, and calculating molecular field using Gauss function[19].

    In the research, the correlation coefficient (q2) and the best principal component value (n) of cross validation were determined by leave one method (LOO) for cross validation. We performed non-cross-validation using previously acquired N values to estimate the general determination factor (r2). In addition, the estimated standard error (SEE) and Fischer statistical values (F) were determined[20]. So far, we constructed the CoMFA and CoMSIA models.

    2.4 3D-QSAR model verification

    2.4.1   Internal verification
    2.4.1.1   Partial least-squares analysis

    PLS verification (cross-verification, non-cross-verification) of 3D-QSAR models was computed in SYBYL X2.0 software. In internal verification, a responsible 3D-QSAR model makes sure to meet the following criteria: q2 > 0.5, r2 > 0.9, the smaller SEE value is better, and the F value is opposite[21-23].

    2.4.1.2   Bootstrap analysis

    After the PLS analysis of 3D-QSAR model, to furtherverify the statistical quality and stability of the model from the training set, the bootstrap analysis for 20 runs was implemented in Sybyl-X2.0 software[24, 25]. And it means the model is more reliable because the rbs2 value is greater than 0.9 and the SDbs value is smaller.

    2.4.1.3   Applicability domain (AD)

    The applicability domain is important to check the stability of model. The applicability field requires that each compound must be within the interval of residual limits (σ±3), and its lever threshold hi must be smaller than critical leverage h*[26, 27].

    $ h_\mathrm{i} = x_\mathrm{i}^{T} (X^{T}X)^{-1}x_\mathrm{i}; h* = 3 × (1 + k)/n $

    where xi (i ≤ n) means the descriptor row-vector of the molecules, X the n × k-1 matrix, n the number of all compounds and k is the number of model descriptors.

    2.4.2   External verification

    Usually, the responsible 3D-QSAR model should need to satisfy both internal and external verification. The external verification of the model is judged based on rm2 > 0.5 in Roy's method; rpred2 > 0.5 in Cruciani-Baroni's method and SDEPext was calulated by the following Eqs[28, 29].

    $ r_{\mathrm{m}}^2=r^2\left(1-\sqrt{\mid r^2-r_o^2} \mid\right) ; r_{\mathrm{pred}}^2=(S D-P R E S S) / S D ; S D E P_{\mathrm{ext}}=\sqrt{\frac{S D}{n_{\mathrm{EXT}}}} $

    2.5 Molecular docking

    The Surflex-Dock function in Sybyl-X2.0 was employed to conduct molecular docking studies to study the binding mode of CDK4/6 kinase and dual inhibitors. The 3D structures of modelling CDK4 (pdb code: 2W9Z, 2.45 Å) and CDK6 (pdb code: 2EUF, 3.00 Å) were used for molecular docking. A suitable docking pocket is generated in the mating mode with default values for adjustable parameters. The active pocket (light gray surface) was defined by amino acids and ranges from about 0.5 Å to the original molecules and the protomols (Fig. 3). The original molecules were re-docked to binding pockets to validate the dependability of docking method. The newly designed compounds and template molecules were docked into CDK4/6. Lastly, the binding models were analyzed using the Analytics Complex tool of PyMOL[30] and Discovery Studio 3.0[31] softwares.

    Figure 3

    Figure 3.  Binding pockets of (a) CDK4 and (b) CDK6
    DownLoad: Full-Size Img PowerPoint

    2.6 MD simulation

    MD simulation was executed with AMBER16 package[32, 33]. Receptor loaded ff99SB force field[34] and ligand loaded general GAFF[35], whole complex system was solvated in TIP3PBOX (Buffer ≥ 10.0 Å) and then the system was electrically neutral by adding sodium and chloride ions.

    We optimized the system to avoid unreasonable structural areas in the system. Then the system was gradually heated to 300 K and ultimately stayed the same. Finally, the temperature fluctuated around 300 K in the next stage. The entire MD process took 2 fs as the time step. Periodic boundary conditions were used to keep constant pressure and temperature. The Langevin dynamics method was used to adjust the temperature at a collision frequency of 2 ps.

    The Particle Mesh Ewald (PME) method was employed to handle longrange electrostatics and the range of actual cutoff spatial interaction was less than 1 nm. Hydrogen atoms were limited by SHAKE method. Next, 100 ns MD simulation was performed, and save their trajectory every 2 ps. The MM/GBSA algorithm was used to process procreant MD simulation trajectories and calculate the binding energy with crystal complexes of different ligands[36, 37]. Total of 100 frames were collected at last 10 ns to calculate the energy. Formula is as below:

    $ \begin{array}{c} \Delta G_{\text {bind }}=G_{\text {complex }}-\left(G_{\text {protein }}+G_{\text {ligand })}\right. \\ \Delta G_{\text {bind }}=\Delta G_{\text {gas }}+\Delta G_{\text {sol }} \\ \Delta G_{\text {gas }}=\Delta E_{\text {ele }}+\Delta E_{\text {vdw}} ; \Delta G_{\text {sol }}=\Delta E_{\text {GB }}+\Delta E_{\text {SURF }} \end{array} $

    2.7 ADME/T property and synthesizable evaluation

    ADME/T, including absorption, distribution, metabolism, elimination and toxicity, is a critical parameter that is commonly used in clinical trials and the selection of the development of drugs[38]. We evaluate the synthetic availability by utilizing SwissADME website (http://www.swissadme.ch/index.php). For uploaded compounds, the web server can predict the percentage of absorption of the drug in intestine of human. When the absorbent degree (HIA) is less than 30%, poor absorption is commonly considered. A higher VD value (VDss logL/kg) indicates that the molecules are mainly distributed in the tissue. Online tools assess whether molecules may be metabolized by CYP450 subtypes, or become CYP450 inhibitors. The total clearance rate (CL tot) is about bio-utilization and the dosing rate is determined to achieve a stable state concentration (logmL/min/kg). Skin sensitivity/AMES toxicity/hepatotoxicity message was made to evaluate the possible toxicity of molecules.

    The synthesizable evaluation of newly designed compounds was calculated by loading the SwissADME website (http://www.swissadme.ch). The scoresyn was closer to 10, the more difficult to synthesize. The value closer to 1 meant the synthesis route is relatively simpler. Finally, we designed a feasible synthesis route according to the existing synthesis methods in relevant articles.

    3. RESULTS AND DISCUSSION

    3.1 3D-QSAR model analysis and statistical verification

    The compounds of training set were modeled by CoMFA and CoMSIA methods and obtain reliable models. We can see that those force fields (Table 2) have made the main contribution to the activity and should be considered in the process of designing new compounds.

    Table 2

    Table 2.  Correlation Coefficients of 3D-QSAR Models
    DownLoad: CSV
    Correlation coefficients CoMFA CDK4 CoMSIACDK4 COMFA CDK6 COMSIA CDK6
    Internal verification
    q2 0.543 0.518 0.624 0.584
    r2 0.967 0.937 0.984 0.975
    F 114.496 49.932 211.615 113.305
    SEE 0.130 0.183 0.105 0.136
    N 8 9 9 10
    rbs2 0.992 0.982 0.997 0.990
    SDbs 0.006 0.012 0.002 0.006
    External verification
    rm2 0.8194 0.5085 0.7067 0.5835
    rpred2 0.7538 0.5339 0.6828 0.6150
    SDEPext 0.2762 0.3801 0.3992 0.4398
    Field contribution
    S 0.771 0.209 0.767 0.184
    E 0.229 0.321 0.233 0.319
    H - 0.470 - 0.497

    In the CoMFACDK4 model, the steric and electrostatic fields contribute 77.1% and 22.9%, respectively, indicating the former is greater.

    In the CoMSIACDK4 model, the steric, electrostatic and hydrophobic fields in the model contribute 20.9%, 32.1% and 47.0%, respectively, indicating that the hydrogen bond donor fields and electrostatic fields contribute greater.

    In the CoMFACDK6 model, the steric and electrostatic fields contribute 76.7% and 23.3%, respectively. It shows that the contribution of the steric field to the model is greater.

    In the CoMSIACDK6 model, the steric, electrostatic and hydrophobic fields in the model contribute 18.4%, 31.9% and 49.7%, respectively. Therefore, the electrostatic and hydrogen bond donor fields have greater contributes.

    Generally, the model needs to satisfy standards of internal and external verification to verify the 3D-QSAR model has a reliable predictive ability. The models all satisfy the standards (Table 2 and Fig. 4), which indicates the models constructed is statistically significant.

    Figure 4

    Figure 4.  Williams plots for 3D-QSAR models
    DownLoad: Full-Size Img PowerPoint

    Finally, by observing the fit of the experimental activity value (pKi) and the predicted activity value (pKi) in Table 3 and Fig. 5, we found that almost all the compounds are close to or located on the trend line of Fig. 5, with a good linear relationship. In Fig. 4, the application domain of the model is characterized by the Williams diagrams, and the test set compounds are all within the application domain of the models. In conclusion, the established 3D-QSAR model has good predictive ability and robustness.

    Table 3

    Table 3.  Predicted Activity Values (pKi) and Residual in the 3D-QSAR Model
    DownLoad: CSV
    No. pKiCDK4 (residual) pKiCDK6 (residual)
    CoMFA CoMSIA CoMFA CoMSIA
    011, 2* 8.862 –0.138 8.797 –0.203 7.882 0.413 8.019 0.550
    021*, 2 8.798 –0.202 8.712 –0.288 8.088 –0.009 8.097 0.000
    031, 2 8.804 0.105 8.791 0.092 8.109 0.063 8.099 0.053
    041, 2 8.817 0.118 8.805 0.106 7.882 –0.077 7.944 –0.015
    051, 2 8.157 –0.065 8.016 –0.206 7.059 0.027 7.065 0.033
    061, 2 8.628 0.230 8.581 0.183 7.536 0.013 7.403 –0.120
    071, 2 8.034 –0.121 7.925 –0.230 7.065 –0.195 7.180 –0.080
    081*, 2 8.769 0.070 8.741 0.042 8.086 0.086 8.106 0.106
    091, 2* 7.950 –0.050 8.050 0.050 7.530 0.021 7.531 0.022
    101, 2 8.227 –0.074 8.312 0.011 7.197 0.017 7.150 –0.030
    111, 2 7.797 0.076 7.834 0.113 6.231 –0.083 6.272 –0.042
    121*, 2 8.092 0.296 8.430 0.634 7.572 0.019 7.557 0.004
    131, 2 8.082 –0.015 8.193 0.096 8.670 –0.029 8.651 –0.048
    141*, 2* 7.919 –0.479 7.775 –0.623 7.874 –0.348 8.725 0.503
    151, 2 7.981 –0.116 7.909 –0.188 8.089 0.130 8.074 0.115
    161, 2 8.947 –0.053 8.701 –0.299 8.495 –0.028 8.333 –0.190
    171, 2 8.673 –0.026 8.809 0.110 8.237 0.015 8.307 0.085
    181, 2* 7.588 0.119 7.806 0.337 7.194 –0.444 7.452 –0.186
    191*, 2 8.527 0.305 8.429 0.207 7.683 0.045 7.746 0.108
    201, 2 8.981 –0.019 8.962 –0.038 7.545 0.036 7.579 0.070
    211, 2* 8.258 0.036 8.214 –0.008 7.251 –0.795 7.400 –0.646
    221, 2* 8.555 0.032 8.640 0.117 7.491 –0.363 8.006 0.152
    231, 2 7.074 –0.018 7.115 0.023 6.166 –0.063 6.114 –0.115
    241, 2 7.206 0.051 7.057 –0.098 7.586 0.017 7.525 –0.044
    251, 2 8.020 0.020 8.088 0.088 5.836 0.059 5.817 0.040
    261, 2 7.338 –0.029 7.212 –0.155 5.797 0.109 6.003 0.315
    271, 2 6.716 –0.029 6.613 –0.132 5.698 –0.066 5.704 –0.060
    281*, 2* 8.250 0.250 8.580 0.580 7.208 –0.330 7.136 –0.402
    291, 2 8.314 0.013 8.268 –0.033 7.587 –0.015 7.380 –0.222
    301*, 2 7.552 0.344 7.780 0.572 6.757 0.077 6.691 0.011
    311, 2* 7.975 0.016 8.091 0.132 6.664 0.490 6.467 0.293
    321, 2 6.538 0.029 6.528 0.019 6.180 0.151 6.032 0.003
    331, 2 8.117 –0.038 8.092 –0.063 7.341 –0.036 7.475 0.098
    341, 2* 6.674 –0.047 6.971 0.250 6.038 0.330 6.572 0.864
    351, 2 7.711 –0.085 7.698 –0.098 7.464 0.020 7.418 –0.026
    361*, 2 8.198 –0.024 8.265 0.043 6.785 –0.158 6.925 –0.018
    371, 2* 8.084 –0.217 8.054 –0.247 6.944 –0.357 6.891 –0.410
    381, 2 9.096 0.096 9.022 0.022 7.439 0.041 7.576 0.178
    391, 2 8.468 0.070 8.422 0.024 7.640 0.145 7.726 0.231
    401, 2 8.447 0.049 8.345 –0.053 7.306 –0.217 7.224 –0.299
    411*, 2 8.340 –0.058 8.214 –0.184 7.052 –0.142 7.054 –0.140
    421, 2 8.381 –0.017 8.451 0.053 7.470 –0.025 7.414 –0.081
    431, 2 8.303 0.002 8.343 0.042 7.665 –0.034 7.762 0.063
    441, 2 7.592 0.160 7.602 0.170 6.650 0.123 6.603 0.076
    451, 2 8.511 –0.012 8.620 0.097 6.550 –0.004 6.678 0.124
    461*, 2 8.487 –0.036 8.599 0.076 7.552 0.057 7.370 –0.125
    471, 2 8.268 –0.130 8.323 –0.075 7.343 –0.055 7.383 –0.015
    481, 2 8.391 –0.308 8.380 –0.319 7.082 –0.178 7.258 –0.002
    491, 2 7.629 0.009 7.547 –0.073 6.588 0.151 6.428 –0.009
    501*, 2* 8.181 0.503 7.938 0.260 7.524 0.272 6.898 –0.354
    511, 2* 7.901 0.378 7.910 0.387 6.794 –0.018 6.757 –0.055
    521*, 2 7.204 0.144 7.268 0.208 6.644 0.013 6.599 –0.032
    1: CDK4 training set; 2: CDK6 training set; *: test set

    Figure 5

    Figure 5.  Fitting curves of the actual and predicted activities of the training and test sets
    DownLoad: Full-Size Img PowerPoint

    3.2 Contour diagrams analysis of the 3D-QSAR

    We studied the counter diagrams of CoMFA and CoMSIA models by using template compound 16. Based on the feasibility of actual synthesis, it could be used for further structural modification of the template compound.

    Figs. 6a, 6c, 6f, and 6h show the all stereo fields of CoMFA and CoMSIA models. The green contour indicates that the activity of compound will increase with more substituent groups, but the yellow contour is unfavorite. The main color contours are distributed around rings D and E using comprehensive analysis. It implies the introduction of a large steric hindrance group or the increase of the number of groups on ring E; The positions of N(18) of ring D can be substituted, so it can be suitable to introduce small groups or not.

    Figure 6

    Figure 6.  Contour diagrams of CoMFA and CoMSIA. CoMFACDK4 (a) stereo (b) electrostatic; CoMSIACDK4 (c) stereo (d) electrostatic (e) hydrophobic field contour diagrams; CoMFACDK6 (f) stereo (g) electrostatic; CoMSIACDK6 (h) stereo (i) electrostatic (j) hydrophobic
    DownLoad: Full-Size Img PowerPoint

    In the electrostatic field, the blue area showed that the activity of the compound increases gradually with the increase of positively charged groups, while the red area is the unfavorite. Some positively charged groups can be introduced on ring B to increase the biological activity of the compound in Figs. 6b, 6d, 6g, and 6i. Conversely, increasing negatively charged groups on position C(19) of ring D could facilitate the activity of the molecule.

    The yellow region (Fig. 6e) indicated that increasing hydrophobic groups can promote activity, and the C(30) and C(33) substituents of ring E are closer to the yellow area in Fig. 6j. In summary, when designing compounds, rings B, D and E are the main group modified region can enhance the activity.

    3.3 Molecular docking

    We used the Surflex-Dock module in the Sybyl-X2.0 software to perform molecular docking. By re-docking ligand in the initial crystal structure back to the active site and then calculating the RMSD 1.143 Å in CDK4 and 1.147 Å in CDK6 (< 2.000 Å), we can see the docking results were reliable[39]. Since compound 16 is the most active, it was regarded as a template to study the interaction after molecular docking. Then, new compounds were designed.

    Surflex-Docking results of compound 16 with CDK4/6 are shown. We observed that these bidentate hydrogen bonds are formed between the Val96/Val101 amino acids and compound 16 of the N(3) and N(7)-H atoms in CDK4/6 kinase, marked with a red dashed line in the results of re-docking and docking novel and template molecules (Figs.7a and 7b). Nowadays, the relevant literature also reports the contribution of these two hydrogen bonds to the inhibition. Hence, the hydrogen bonds mentioned above should be considered in the new design of compounds.

    Figure 7

    Figure 7.  Re-docked results of the original ligand; Original conformation (yellow) and the re-docked conformation (cyan); Hydrogen bond interaction are marked with red dashed lines
    DownLoad: Full-Size Img PowerPoint

    3.4 Design new compounds and predict activity

    Lead compound 16 was modified in view of the 3D-QSAR models and related results. Subsequently, new inhibitors with higher activity were designed.

    In this article, we designed 10 new compounds and used the 3D-QSAR model to predict their activity, while we also performed molecular docking to assess their binding ability with the target protein by total score. Table 4 lists the designed molecular structure, predicted activity and docking score. Most newly designed compounds have higher pKi and docking scores.

    Table 4

    Table 4.  Information of Template Compound and the New Compounds
    DownLoad: CSV
    No. R1 R3 R4 pKiCDK4 pKiCDK6 Total score
    CoMFA CoMSIA CoMFA CoMSIA CDK4 CDK6
    16 F 8.947 8.701 8.495 8.333 6.569 6.734
    D1 CN 9.586 9.067 8.693 8.614 7.016 7.048
    D2 CN 9.407 9.089 8.627 9.166 6.883 6.956
    D3 CN 9.678 9.692 8.661 9.487 6.727 7.012
    D4 CN 9.543 8.659 8.595 8.895 6.808 7.083
    D5 CN 9.487 9.582 8.383 9.198 6.673 6.960
    D6 CN 9.35 8.966 8.594 8.343 7.031 6.972
    D7 CN 9.523 10.046 8.535 8.273 7.106 7.012
    D8 CN 9.628 10.191 8.616 8.423 6.776 6.914
    D9 CN 9.432 10.167 8.515 8.913 7.173 6.838
    D10 CN 9.662 10.784 8.512 9.630 7.183 7.140
    R2 all: methyl

    By docking the template molecule and 10 new molecules into the active site, we found that these compounds are well incorporated in the active sites (Fig. 8).

    Figure 8

    Figure 8.  Docking conformation of the CDK4/6 crystal structure with the novel and template molecules
    DownLoad: Full-Size Img PowerPoint

    3.5 MD simulation and MM/GBSA

    To validate whether new inhibitors predicted by the 3D-QSAR models and molecular docking can bind well to the CDK4/6 proteins, we performed MD simulations of compounds 16 and D10 with the best average predictive activity value and the best docking scores. The RMSD of CDK4 complexed compound 16/D10 fluctuated around ~4.25 and ~3.75 Å, respectively after 70 ns simulation (Fig. 9a). For the docking results of CDK6 and compound 16/D10, the RMSD of the complexes also remained stable at ~4.0 and ~3.5 Å after 70 ns simulation (Fig. 9b). The RMSDs of above complexes were stable in the late simulation stage. The above analysis results indicated that these four complexes reached a metastable stage, and RMSD values of the designed compound D10 in the stable phase are all smaller than the template molecule, indicating that the designed molecule has better binding ability with the target proteins.

    Figure 9

    Figure 9.  RMSD of complexes CDK4/6 docking with compound 16/D10 for 100 ns simulations
    DownLoad: Full-Size Img PowerPoint

    Then as shown in Fig. 10, we compared the interaction before and after MD simulation. Compounds 16 and D10 both formed hydrogen bonds with amino acid residues Val96 and Lys22 of CDK4, marked with a red dashed line. Before MD simulation, the unstable hydrogen bond formed between compound 16 and lys35 of CDK4 but lys35 changes to hydrogen bond to π-NH interaction, which is marked with magenta dotted line. Similarly, compounds 16 and D10 both formed hydrogen bonds with amino acid residues Val101 of CDK6. Compared with compound 16, compound D10 could form stable hydrogen key with amino acid residues His100 of CDK4/6. Compound D10 also formed hydrogen bond (Lys43) in CDK6 protein, which enhanced the binding between proteins and small molecules to a certain extent.

    Figure 10

    Figure 10.  Interaction between CDK4/6 and compound 16/D10 before and after 100 ns MD simulation
    DownLoad: Full-Size Img PowerPoint

    Furthermore, it has been reported that above hydrogen bonds, Val96 (CDK4)[10], Val101 (CDK6) and His100 (CDK6)[9, 10] can greatly affect the activity of inhibitors, and build a further theoretical basis for the design of inhibitors. The re-docking result of original ligand in the crystal structure also proves the above-mentioned hydrogen bonding interactions.

    After MD simulation, we used MM/GBSA method to decompose binding free energy to find the key residues in the binding process of CDK4/6 and the inhibitor. The binding free energies and decomposed energy items in the MD process are shown and listed in Fig. 11 and Table 5. Compared with compound 16, compound D10 has better binding energy and forms a key residue role with more amino acids, indicating compound D10 may have better bioactivity.

    Figure 11

    Figure 11.  Contributions of key residues of CDK4/6
    DownLoad: Full-Size Img PowerPoint

    Table 5

    Table 5.  Binding Free Energies (kcal/mol)
    DownLoad: CSV
    MM/GBSA CDK4 CDK6
    16 D10 16 D10
    ΔEvdw –59.958 –65.693 –47.138 –56.024
    ΔEele –28.939 –29.910 –14.561 –23.101
    ΔEGB 46.510 49.531 33.663 39.822
    ΔESURF –6.8679 –7.486 –5.4652 –6.380
    ΔGBind –49.2561 –53.558 –33.501 –45.684

    3.6 Synthesizable evaluation and reference synthetic route

    Then synthesizable evaluation scores (Scoresyn) of newly designed compounds ranged between 4.29 and 5.53 which are close to that of the template compound 16, thus proving we could increase the inhibitory activity of compounds by introducing related groups. Meanwhile, introducing related chemical substituents did not make the synthesis process difficult and they all had the appropriate synthesis possibilities (Table 6).

    Table 6

    Table 6.  Synthesis Accessibility of New Compounds
    DownLoad: CSV
    No. SMILES string Scoresyn
    16 Cc1nc([s]c1c2nc(ncc2[F])Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CCCC5 4.17
    D1 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CCCC5 4.29
    D2 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)N[C@@]5([H])CCCC5(C)C 4.98
    D3 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)N[C@@]5([H])CCC[C@]5([H])C(C)(C)[H] 5.37
    D4 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)N[C@]5([H])CNCC5 4.74
    D5 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CCC([F])([F])CC5 4.43
    D6 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CC=CC5 4.77
    D7 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Cl])([H])CC4)N[C@]5([H])CCCC5 4.66
    D8 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Br])([H])CC4)N[C@]5([H])CCCC5 4.72
    D9 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Br])([H])CC4)N[C@@]5([H])CCCC5(C)C 5.22
    D10 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Br])([H])CC4)N[C@@]5([H])CCC[C@]5([H])C(C)(C)[H] 5.53
    Scoresyn: synthetic accessibility from 1(very easy) to 10 (very difficult)

    According to the existing literature[9-11], we designed a more reliable synthesis route (Fig. 12), and proved the CAS number of the required reaction substrate (Table 7), providing a reliable source of synthetic materials and further synthesis possibilities for the synthesis of the newly designed compound (Please refer to the above literatures for more detailed synthetic experiments).

    Figure 12

    Figure 12.  Reference synthesis route of the novel compounds

    Reagents and conditions: (a) CS2, NEt3, Boc2O, DMAP, EtOH, rt, 30 min, ice bath, 15 min, rt, 30 min; (b) NH3, H2O, MeOH, rt, 24 h; (c) ethyl-2-chloroacetoacetate, DMAP, CH2Cl2, rt, 4 h; (d) di-tert-butyl dicarbonate, DCM, Et3N, 8 h, rt; (e) LiNiPr2, CH3CN, –78 ℃, 10 min; o/n, HCl, H2O, 48%; (f) DMF-DMA, reflux, 8 h; (g) 5-bromo-2-nitropyridine, DMSO, Et3N, 120 ℃, 16 h; (h) H2, 10%Pd/C, CH3OH, 6.5~12 h, rt; (i) N, N΄-bis-Boc-S-methylisothiourea, Et3N, HgCl2, DCM, 0 ℃, 0.5 h, rt, 14 h; (j) TFA/DCM/H2O (18:9:1), rt, 16 h; (k) NaOH, 2-methoxyethanol, microwave, 180 ℃, 1 h

    DownLoad: Full-Size Img PowerPoint

    Table 7

    Table 7.  CAS Number of Substrates
    DownLoad: CSV
    D1 D2 D3 D4 D5
    H-A1 1003-03-8 345658-02-8 90226-30-5 79286-79-6 458566-84-2
    D6 D7 D8 D9 D10
    91469-55-5 1003-03-8 1003-03-8 345658-02-8 90226-30-5
    H-A2 109-01-3 109-01-3 109-01-3 109-01-3 109-01-3
    D6 D7 D8 D9 D10
    109-01-3 5382-18-3 54288-70-9 54288-70-9 54288-70-9

    3.7 ADMET profiles

    ADME/T profiles were predicted using pkCSM web server (Table 8) to investigate the druggability of the newly designed compounds. HIA of them changed from 88.371% to 95.057% (> 30%), explaining all newly compounds have high absorptive capacity. The VDs of most new compounds were less than compound 16 (0.71 low ~ 2.81 high), which meant all compounds fall within the proper range. For pharmacokinetics, 3A4 subtypes might metabolize all compounds, and 2D6 subtypes might not metabolize. Additionally, the above molecules could not restrain CYP450 1A2 and 2D6, whereas the others could restrain the CYP450 3A4 subtype. Renal and hepatic tissues could be clear based on total clearance rate. The predicted toxicity results show the lead compound and all newly designed compounds may damage the human liver, but are not sensitizing to the skin. In summary, we considered newly designed compounds could be as potential inhibitors by ADMET.

    Table 8

    Table 8.  ADMET Properties
    DownLoad: CSV
    No. Abs Dis Met CYP (substrate) Met CYP (inhibitor) Ex Tox
    HIA VDss 2D6 3A4 1A2 2D6 3A4 CL tot AMES H-HT Skin Sen
    16 93.138 1.170 No Yes No No No 0.637 No Yes No
    D1 84.135 1.175 No Yes No No No 0.602 No Yes No
    D2 85.241 1.245 No Yes No No Yes 0.437 No Yes No
    D3 85.477 1.382 No Yes No No Yes 0.210 No Yes No
    D4 73.096 1.587 No Yes No No No 0.560 No Yes No
    D5 85.174 1.097 No Yes No No No 0.441 No Yes No
    D6 83.283 1.168 No Yes No No No 0.557 No Yes No
    D7 93.770 0.783 No Yes No No Yes 0.417 No Yes No
    D8 93.715 0.794 No Yes No No Yes 0.364 No Yes No
    D9 94.821 0.871 No Yes No No No 0.319 No Yes No
    D10 95.057 1.005 No Yes No No Yes 0.274 No Yes No

    4. CONCLUSION

    In this article, the reliable 3D-QSAR model was constructed as CDK4/6 dual inhibitors. Compared with CoMFA model, CoMSIA model introduces hydrophobic field, hydrogen bond donor field and hydrogen bond acceptor field to improve the accuracy of model. 3D-QSAR results showed that the model satisfied relevant statistical correlation and strict internal and external verification requirements, indicating the model has good predictive ability. At the same time, molecular docking and MD simulation further studied the newly designed compounds. MD simulation showed that relevant hydrogen bond and π-NH interactions were important to the binding process. We designed new compounds with good bioactivity, good ADME/T properties, suitable synthesis possibility and reference synthesis process based on 3D-QSAR model. Besides, compounds 16 and D10 could bind to CDK4/6 by hydrogen key and π-alkyl interaction. There are some key residues Lys22/Lys35/Val96 (CDK4) and Lys43/His100/Val101 (CDK6) in the binding sites. These results showed compound D10 may have better bioactivity and provide a theoretical foundation for the further research. Therefore, we provide here a useful reference and new idea for further study and synthesis of dual CDK4/6 inhibitors by using reliable computer-aided drug design methods.


    1. [1]

      Shapiro, G. I. Cyclin-dependent kinase pathways as targets for cancer treatment. J. Clin. Oncol. 2006, 24, 1770–1783. doi: 10.1200/JCO.2005.03.7689

    2. [2]

      Bartek, J.; Bartkova, J.; Lukas, J. The retinoblastoma protein pathway in cell cycle control and cancer. Exp. Cell. Res. 1997, 237, 1–6. doi: 10.1006/excr.1997.3776

    3. [3]

      Nevins, J. R. The Rb/E2F pathway and cancer. Hum. Mol. Genet. 2001, 10, 699–703. doi: 10.1093/hmg/10.7.699

    4. [4]

      Griggs, J. J.; Wolff, A. C. Cyclin-dependent kinase 4/6 inhibitors in the treatment of breast cancer: More breakthroughs and an embarrassment of riches. J. Clin. Oncol. 2017, 35, 2857–2859. doi: 10.1200/JCO.2017.73.9375

    5. [5]

      Boér, K. Impact of palbociclib combinations on treatment of advanced estrogen receptor-positive/human epidermal growth factor 2-negative breast cancer. Onco. Targets. Ther. 2016, 9, 6119–6125. doi: 10.2147/OTT.S77033

    6. [6]

      Cunningham, N. C.; Turner, N. C. Understanding divergent trial results of adjuvant CDK4/6 inhibitors for early stage breast cancer. Cancer Cell 2021, 39, 307–309. doi: 10.1016/j.ccell.2021.02.011

    7. [7]

      Kwapisz, D. Cyclin-dependent kinase 4/6 inhibitors in breast cancer: palbociclib, ribociclib, and abemaciclib. Breast Cancer Res. Treat. 2017, 166, 41–54. doi: 10.1007/s10549-017-4385-3

    8. [8]

      Shan, H.; Ma, X.; Yan, G.; Luo, M.; Zhong, X.; Lan, S.; Yang, J.; Liu, Y.; Pu, C.; Tong, Y. Discovery of a novel covalent CDK4/6 inhibitor based on palbociclib scaffold. Eur. J. Med. Chem. 2021, 219, 113432–113440. doi: 10.1016/j.ejmech.2021.113432

    9. [9]

      Tadesse, S.; Zhu, G.; Mekonnen, L. B.; Jimma, L. L.; Yu, M.; Brown, M. P.; Wang, S. A novel series of N-(pyridin-2-yl)-4-(thiazol-5-yl)pyrimidin-2-amines as highly potent CDK4/6 inhibitors. Future Med. Chem. 2017, 13, 1495–1506.

    10. [10]

      Tadesse, S.; Yu, M.; Mekonnen, L. B.; Lam, F.; Islam, S.; Tomusange, K.; Rahaman, M. H.; Noll, B.; Basnet, S. K. C.; Teo, T. Highly potent, selective, and orally bioavailable 4-thiazol-n-(pyridin-2-yl)pyrimidin-2-amine cyclin-dependent kinases 4 and 6 inhibitors as anticancer drug candidates: Design, synthesis, and evaluation. J. Med. Chem. 2017, 60, 1892–1915. doi: 10.1021/acs.jmedchem.6b01670

    11. [11]

      Tadesse, S.; Bantie, L.; Tomusange, K.; Yu, M.; Islam, S.; Bykovska, N.; Noll, B.; Zhu, G.; Li, P.; Lam, F. Discovery and pharmacological characterization of a novel series of highly selective inhibitors of cyclin-dependent kinases 4 and 6 as anticancer agents. Brit. J. Pharmacol. 2018, 175, 2399–2413. doi: 10.1111/bph.13974

    12. [12]

      John, E. B. Sybyl-X, Molecular modeling software packages, Version 2.0. TRIPOS, Associates, Inc., St. Louis, MO, USA 2012.

    13. [13]

      Fu, L.; Chen, Y.; Xu, C. M.; Wu, T.; Guo, H. M.; Lin, Z. H.; Wang, R., Shu, M. 3D-QSAR, HQSAR, molecular docking, and new compound design study of 1, 3, 6-trisubstituted 1, 4-diazepan-7-ones as human KLK7 inhibitors. Med. Chem. Res. 2020, 29, 1012–1029. doi: 10.1007/s00044-020-02542-3

    14. [14]

      Romero-Parra, J.; Chung, H.; Tapia, R. A.; Faúndez, M.; Morales-Verdejo, C.; Lorca, M.; Lagos, C. F.; Di Marzo, V.; David Pessoa-Mahana, C.; Mella, J. Combined CoMFA and CoMSIA 3D-QSAR study of benzimidazole and benzothiophene derivatives with selective affinity for the CB2 cannabinoid receptor. Eur. J. Pharm. Sci. 2017, 101, 1–10. doi: 10.1016/j.ejps.2017.01.037

    15. [15]

      Manouchehrizadeh, E.; Mostoufi, A.; Tahanpesar, E.; Fereidoonnezhad, M. Alignment-independent 3D-QSAR and molecular docking studies of tacrine-4-oxo-4H-Chromene hybrids as anti-Alzheimer's agents. Comput. Biol. Chem. 2019, 80, 463–471. doi: 10.1016/j.compbiolchem.2019.05.010

    16. [16]

      Fu, L.; Chen, Y.; Guo, H. M.; Xu, L.; Tan, M. N.; Dong, Y.; Shu, M.; Wang, R.; Lin, Z. H A selectivity study of polysubstituted pyridinylimidazoles as dual inhibitors of JNK3 and p38α MAPK based on 3D-QSAR, molecular docking, and molecular dynamics simulation. Struct. Chem. 2021, 32, 819–834. doi: 10.1007/s11224-020-01668-9

    17. [17]

      Farrugia, L. J. WinGX and ORTEP for Windows: An update. J. Appl. Crystallogr. 2012, 45, 849–854. doi: 10.1107/S0021889812029111

    18. [18]

      Clark, M.; Cramer, R. D.; Jones, D. M.; Patterson, D. E.; Simeroth, P. E. Comparative molecular field analysis (CoMFA). 2. Toward its use with 3D-structural databases. Tetrahedron Comput. Methodol. 1990, 3, 47–59. doi: 10.1016/0898-5529(90)90120-W

    19. [19]

      Klebe, G.; Abraham, U. Comparative molecular similarity index analysis (CoMSIA) to study hydrogen-bonding properties and to score combinatorial libraries. J. Comput. Aid. Mol. Des. 1999, 13, 1–10. doi: 10.1023/A:1008047919606

    20. [20]

      Bush, B.; Nachbar, R. J. Sample-distance partial least squares: PLS optimized for many variables, with application to CoMFA. J. Comput. Aid. Mol. Des. 1993, 7, 587–619. doi: 10.1007/BF00124364

    21. [21]

      Wendt, B.; Cramer, R. Challenging the gold standard for 3D-QSAR: template CoMFA versus X-ray alignment. J. Comput. Aid. Mol. Des. 2014, 28, 803–824. doi: 10.1007/s10822-014-9761-z

    22. [22]

      Gu, C.; Goodarzi, M.; Yang, X.; Bian, Y.; Sun, C.; Jiang, X. Predictive insight into the relationship between AhR binding property and toxicity of polybrominated diphenyl ethers by PLS-derived QSAR. Toxicol. Lett. 2012, 208, 269–274. doi: 10.1016/j.toxlet.2011.11.010

    23. [23]

      Jian-Feng, L.; Li-Min, L. Structural characterization and aquatic toxicity prediction of esters. Chin. J. Struct. Chem. 2021, 40, 711–721.

    24. [24]

      Jing, P.; Zhao, S.; Ruan, S.; Sui, Z.; Chen, L.; Jiang, L.; Qian, B. Quantitative studies on structure–ORAC relationships of anthocyanins from eggplant and radish using 3D-QSAR. Food Chem. 2014, 145, 365–371. doi: 10.1016/j.foodchem.2013.08.082

    25. [25]

      Lei, Z.; Keng-Chang, T.; Lupei, D.; Hao, F.; Minyong, L.; Wenfang, X. How to generate reliable and predictive CoMFA models. Curr. Med. Chem. 2011, 18, 923–930. doi: 10.2174/092986711794927702

    26. [26]

      Kar, S.; Roy, K.; Leszczynski, J. Applicability domain: a step toward confident predictions and decidability for QSAR modeling. Comput. Toxicol. 2018, 1800, 141–169.

    27. [27]

      Tian, Y.; Zhang, S.; Yin, H.; Yan, A. Quantitative structure-activity relationship (QSAR) models and their applicability domain analysis on HIV-1 protease inhibitors by machine learning methods. Chemometr. Intell. Lab. 2020, 196, 103888–103892. doi: 10.1016/j.chemolab.2019.103888

    28. [28]

      Cruciani, G.; Baroni, M.; Clementi, S.; Costantino, G.; Riganelli, D.; Skagerberg, B. Predictive ability of regression models. Part I: Standard deviation of prediction errors (SDEP). J. Chemometr. 1992, 6, 335–346 doi: 10.1002/cem.1180060604

    29. [29]

      Tosco, P.; Balle, T. A 3D-QSAR-driven approach to binding mode and affinity prediction. J. Chem. Inf. Model. 2012, 52, 302–307. doi: 10.1021/ci200411s

    30. [30]

      Warren, L. D. The PyMOL Molecular Graphics System, Version 2.4, Schrödinger, LLC., New York, USA 2020.

    31. [31]

      Jason, B. Discovery Studio Modelling Environment, Version 3.0, Accelrys Software, Inc., San Diego, CA 92121, USA 2012.

    32. [32]

      Götz, A. W.; Williamson, M. J.; Xu, D.; Poole, D.; Le Grand, S.; Walker, R. C. Routine microsecond molecular dynamics simulations with AMBER on GPUs. 1. Generalized born. J. Chem. Theory Comput. 2012, 8, 1542–1555. doi: 10.1021/ct200909j

    33. [33]

      Salomon-Ferrer, R.; Götz, A. W.; Poole, D.; Le Grand, S.; Walker, R. C. Routine microsecond molecular dynamics simulations with AMBER on GPUs. 2. Explicit solvent particle mesh Ewald. J. Chem. Theory Comput. 2013, 9, 3878–3888. doi: 10.1021/ct400314y

    34. [34]

      Sprenger, K. G.; Jaeger, V. W.; Pfaendtner, J. The general AMBER force field (GAFF) can accurately predict thermodynamic and transport properties of many ionic liquids. J. Phys. Chem. B 2015, 119, 5882–5895. doi: 10.1021/acs.jpcb.5b00689

    35. [35]

      Lindorff-Larsen, K.; Piana, S.; Palmo, K.; Maragakis, P.; Klepeis, J. L.; Dror, R. O.; Shaw, D. E. Improved side-chain torsion potentials for the Amber ff99SB protein force field. Proteins 2010, 78, 1950–1958. doi: 10.1002/prot.22711

    36. [36]

      Sun, H.; Duan, L.; Chen, F.; Liu, H.; Wang, Z.; Pan, P.; Zhu, F.; Zhang, J. Z. H.; Hou, T. Assessing the performance of MM/PBSA and MM/GBSA methods. 7. Entropy effects on the performance of end-point binding free energy calculation approaches. Phys. Chem. Chem. Phys. 2018, 20, 14450–14460. doi: 10.1039/C7CP07623A

    37. [37]

      Huang, K.; Luo, S.; Cong, Y.; Zhong, S.; Zhang, J. Z. H.; Duan, L. An accurate free energy estimator: based on MM/PBSA combined with interaction entropy for protein-ligand binding affinity. Nanoscale 2020, 12, 10737–10750. doi: 10.1039/C9NR10638C

    38. [38]

      Zekri, A.; Harkati, D.; Kenouche, S.; Saleh, B. A. QSAR modeling, docking, ADME and reactivity of indazole derivatives as antagonizes of estrogen receptor alpha (ER-α) positive in breast cancer. J. Mol. Struct. 2020, 1217, 128442–128448. doi: 10.1016/j.molstruc.2020.128442

    39. [39]

      Abdizadeh, T.; Ghodsi, R.; Hadizadeh, F. 3D-QSAR (CoMFA, CoMSIA) and molecular docking studies on histone deacetylase 1 selective inhibitors. Recent Pat. Anti-Canc. 2017, 12, 365–383.

  • Figure 1  Structures of CDK4/6 inhibitors

    下载: 全尺寸图片 幻灯片

    Figure 2  (a) Alignment of dual CDK4/6 inhibitors; (b) Structure of compound 16

    下载: 全尺寸图片 幻灯片

    Figure 3  Binding pockets of (a) CDK4 and (b) CDK6

    下载: 全尺寸图片 幻灯片

    Figure 4  Williams plots for 3D-QSAR models

    下载: 全尺寸图片 幻灯片

    Figure 5  Fitting curves of the actual and predicted activities of the training and test sets

    下载: 全尺寸图片 幻灯片

    Figure 6  Contour diagrams of CoMFA and CoMSIA. CoMFACDK4 (a) stereo (b) electrostatic; CoMSIACDK4 (c) stereo (d) electrostatic (e) hydrophobic field contour diagrams; CoMFACDK6 (f) stereo (g) electrostatic; CoMSIACDK6 (h) stereo (i) electrostatic (j) hydrophobic

    下载: 全尺寸图片 幻灯片

    Figure 7  Re-docked results of the original ligand; Original conformation (yellow) and the re-docked conformation (cyan); Hydrogen bond interaction are marked with red dashed lines

    下载: 全尺寸图片 幻灯片

    Figure 8  Docking conformation of the CDK4/6 crystal structure with the novel and template molecules

    下载: 全尺寸图片 幻灯片

    Figure 9  RMSD of complexes CDK4/6 docking with compound 16/D10 for 100 ns simulations

    下载: 全尺寸图片 幻灯片

    Figure 10  Interaction between CDK4/6 and compound 16/D10 before and after 100 ns MD simulation

    下载: 全尺寸图片 幻灯片

    Figure 11  Contributions of key residues of CDK4/6

    下载: 全尺寸图片 幻灯片

    Figure 12  Reference synthesis route of the novel compounds

    Reagents and conditions: (a) CS2, NEt3, Boc2O, DMAP, EtOH, rt, 30 min, ice bath, 15 min, rt, 30 min; (b) NH3, H2O, MeOH, rt, 24 h; (c) ethyl-2-chloroacetoacetate, DMAP, CH2Cl2, rt, 4 h; (d) di-tert-butyl dicarbonate, DCM, Et3N, 8 h, rt; (e) LiNiPr2, CH3CN, –78 ℃, 10 min; o/n, HCl, H2O, 48%; (f) DMF-DMA, reflux, 8 h; (g) 5-bromo-2-nitropyridine, DMSO, Et3N, 120 ℃, 16 h; (h) H2, 10%Pd/C, CH3OH, 6.5~12 h, rt; (i) N, N΄-bis-Boc-S-methylisothiourea, Et3N, HgCl2, DCM, 0 ℃, 0.5 h, rt, 14 h; (j) TFA/DCM/H2O (18:9:1), rt, 16 h; (k) NaOH, 2-methoxyethanol, microwave, 180 ℃, 1 h

    下载: 全尺寸图片 幻灯片

    Table 1.  Chemical Structures of Dual CDK4/6 Inhibitors
    下载: 导出CSV

    Table 2.  Correlation Coefficients of 3D-QSAR Models

    Correlation coefficients CoMFA CDK4 CoMSIACDK4 COMFA CDK6 COMSIA CDK6
    Internal verification
    q2 0.543 0.518 0.624 0.584
    r2 0.967 0.937 0.984 0.975
    F 114.496 49.932 211.615 113.305
    SEE 0.130 0.183 0.105 0.136
    N 8 9 9 10
    rbs2 0.992 0.982 0.997 0.990
    SDbs 0.006 0.012 0.002 0.006
    External verification
    rm2 0.8194 0.5085 0.7067 0.5835
    rpred2 0.7538 0.5339 0.6828 0.6150
    SDEPext 0.2762 0.3801 0.3992 0.4398
    Field contribution
    S 0.771 0.209 0.767 0.184
    E 0.229 0.321 0.233 0.319
    H - 0.470 - 0.497
    下载: 导出CSV

    Table 3.  Predicted Activity Values (pKi) and Residual in the 3D-QSAR Model

    No. pKiCDK4 (residual) pKiCDK6 (residual)
    CoMFA CoMSIA CoMFA CoMSIA
    011, 2* 8.862 –0.138 8.797 –0.203 7.882 0.413 8.019 0.550
    021*, 2 8.798 –0.202 8.712 –0.288 8.088 –0.009 8.097 0.000
    031, 2 8.804 0.105 8.791 0.092 8.109 0.063 8.099 0.053
    041, 2 8.817 0.118 8.805 0.106 7.882 –0.077 7.944 –0.015
    051, 2 8.157 –0.065 8.016 –0.206 7.059 0.027 7.065 0.033
    061, 2 8.628 0.230 8.581 0.183 7.536 0.013 7.403 –0.120
    071, 2 8.034 –0.121 7.925 –0.230 7.065 –0.195 7.180 –0.080
    081*, 2 8.769 0.070 8.741 0.042 8.086 0.086 8.106 0.106
    091, 2* 7.950 –0.050 8.050 0.050 7.530 0.021 7.531 0.022
    101, 2 8.227 –0.074 8.312 0.011 7.197 0.017 7.150 –0.030
    111, 2 7.797 0.076 7.834 0.113 6.231 –0.083 6.272 –0.042
    121*, 2 8.092 0.296 8.430 0.634 7.572 0.019 7.557 0.004
    131, 2 8.082 –0.015 8.193 0.096 8.670 –0.029 8.651 –0.048
    141*, 2* 7.919 –0.479 7.775 –0.623 7.874 –0.348 8.725 0.503
    151, 2 7.981 –0.116 7.909 –0.188 8.089 0.130 8.074 0.115
    161, 2 8.947 –0.053 8.701 –0.299 8.495 –0.028 8.333 –0.190
    171, 2 8.673 –0.026 8.809 0.110 8.237 0.015 8.307 0.085
    181, 2* 7.588 0.119 7.806 0.337 7.194 –0.444 7.452 –0.186
    191*, 2 8.527 0.305 8.429 0.207 7.683 0.045 7.746 0.108
    201, 2 8.981 –0.019 8.962 –0.038 7.545 0.036 7.579 0.070
    211, 2* 8.258 0.036 8.214 –0.008 7.251 –0.795 7.400 –0.646
    221, 2* 8.555 0.032 8.640 0.117 7.491 –0.363 8.006 0.152
    231, 2 7.074 –0.018 7.115 0.023 6.166 –0.063 6.114 –0.115
    241, 2 7.206 0.051 7.057 –0.098 7.586 0.017 7.525 –0.044
    251, 2 8.020 0.020 8.088 0.088 5.836 0.059 5.817 0.040
    261, 2 7.338 –0.029 7.212 –0.155 5.797 0.109 6.003 0.315
    271, 2 6.716 –0.029 6.613 –0.132 5.698 –0.066 5.704 –0.060
    281*, 2* 8.250 0.250 8.580 0.580 7.208 –0.330 7.136 –0.402
    291, 2 8.314 0.013 8.268 –0.033 7.587 –0.015 7.380 –0.222
    301*, 2 7.552 0.344 7.780 0.572 6.757 0.077 6.691 0.011
    311, 2* 7.975 0.016 8.091 0.132 6.664 0.490 6.467 0.293
    321, 2 6.538 0.029 6.528 0.019 6.180 0.151 6.032 0.003
    331, 2 8.117 –0.038 8.092 –0.063 7.341 –0.036 7.475 0.098
    341, 2* 6.674 –0.047 6.971 0.250 6.038 0.330 6.572 0.864
    351, 2 7.711 –0.085 7.698 –0.098 7.464 0.020 7.418 –0.026
    361*, 2 8.198 –0.024 8.265 0.043 6.785 –0.158 6.925 –0.018
    371, 2* 8.084 –0.217 8.054 –0.247 6.944 –0.357 6.891 –0.410
    381, 2 9.096 0.096 9.022 0.022 7.439 0.041 7.576 0.178
    391, 2 8.468 0.070 8.422 0.024 7.640 0.145 7.726 0.231
    401, 2 8.447 0.049 8.345 –0.053 7.306 –0.217 7.224 –0.299
    411*, 2 8.340 –0.058 8.214 –0.184 7.052 –0.142 7.054 –0.140
    421, 2 8.381 –0.017 8.451 0.053 7.470 –0.025 7.414 –0.081
    431, 2 8.303 0.002 8.343 0.042 7.665 –0.034 7.762 0.063
    441, 2 7.592 0.160 7.602 0.170 6.650 0.123 6.603 0.076
    451, 2 8.511 –0.012 8.620 0.097 6.550 –0.004 6.678 0.124
    461*, 2 8.487 –0.036 8.599 0.076 7.552 0.057 7.370 –0.125
    471, 2 8.268 –0.130 8.323 –0.075 7.343 –0.055 7.383 –0.015
    481, 2 8.391 –0.308 8.380 –0.319 7.082 –0.178 7.258 –0.002
    491, 2 7.629 0.009 7.547 –0.073 6.588 0.151 6.428 –0.009
    501*, 2* 8.181 0.503 7.938 0.260 7.524 0.272 6.898 –0.354
    511, 2* 7.901 0.378 7.910 0.387 6.794 –0.018 6.757 –0.055
    521*, 2 7.204 0.144 7.268 0.208 6.644 0.013 6.599 –0.032
    1: CDK4 training set; 2: CDK6 training set; *: test set
    下载: 导出CSV

    Table 4.  Information of Template Compound and the New Compounds

    No. R1 R3 R4 pKiCDK4 pKiCDK6 Total score
    CoMFA CoMSIA CoMFA CoMSIA CDK4 CDK6
    16 F 8.947 8.701 8.495 8.333 6.569 6.734
    D1 CN 9.586 9.067 8.693 8.614 7.016 7.048
    D2 CN 9.407 9.089 8.627 9.166 6.883 6.956
    D3 CN 9.678 9.692 8.661 9.487 6.727 7.012
    D4 CN 9.543 8.659 8.595 8.895 6.808 7.083
    D5 CN 9.487 9.582 8.383 9.198 6.673 6.960
    D6 CN 9.35 8.966 8.594 8.343 7.031 6.972
    D7 CN 9.523 10.046 8.535 8.273 7.106 7.012
    D8 CN 9.628 10.191 8.616 8.423 6.776 6.914
    D9 CN 9.432 10.167 8.515 8.913 7.173 6.838
    D10 CN 9.662 10.784 8.512 9.630 7.183 7.140
    R2 all: methyl
    下载: 导出CSV

    Table 5.  Binding Free Energies (kcal/mol)

    MM/GBSA CDK4 CDK6
    16 D10 16 D10
    ΔEvdw –59.958 –65.693 –47.138 –56.024
    ΔEele –28.939 –29.910 –14.561 –23.101
    ΔEGB 46.510 49.531 33.663 39.822
    ΔESURF –6.8679 –7.486 –5.4652 –6.380
    ΔGBind –49.2561 –53.558 –33.501 –45.684
    下载: 导出CSV

    Table 6.  Synthesis Accessibility of New Compounds

    No. SMILES string Scoresyn
    16 Cc1nc([s]c1c2nc(ncc2[F])Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CCCC5 4.17
    D1 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CCCC5 4.29
    D2 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)N[C@@]5([H])CCCC5(C)C 4.98
    D3 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)N[C@@]5([H])CCC[C@]5([H])C(C)(C)[H] 5.37
    D4 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)N[C@]5([H])CNCC5 4.74
    D5 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CCC([F])([F])CC5 4.43
    D6 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CCN(C)CC4)NC5([H])CC=CC5 4.77
    D7 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Cl])([H])CC4)N[C@]5([H])CCCC5 4.66
    D8 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Br])([H])CC4)N[C@]5([H])CCCC5 4.72
    D9 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Br])([H])CC4)N[C@@]5([H])CCCC5(C)C 5.22
    D10 Cc1nc([s]c1c2nc(ncc2C#N)Nc3ncc(cc3)N4CC[C@@]([Br])([H])CC4)N[C@@]5([H])CCC[C@]5([H])C(C)(C)[H] 5.53
    Scoresyn: synthetic accessibility from 1(very easy) to 10 (very difficult)
    下载: 导出CSV

    Table 7.  CAS Number of Substrates

    D1 D2 D3 D4 D5
    H-A1 1003-03-8 345658-02-8 90226-30-5 79286-79-6 458566-84-2
    D6 D7 D8 D9 D10
    91469-55-5 1003-03-8 1003-03-8 345658-02-8 90226-30-5
    H-A2 109-01-3 109-01-3 109-01-3 109-01-3 109-01-3
    D6 D7 D8 D9 D10
    109-01-3 5382-18-3 54288-70-9 54288-70-9 54288-70-9
    下载: 导出CSV

    Table 8.  ADMET Properties

    No. Abs Dis Met CYP (substrate) Met CYP (inhibitor) Ex Tox
    HIA VDss 2D6 3A4 1A2 2D6 3A4 CL tot AMES H-HT Skin Sen
    16 93.138 1.170 No Yes No No No 0.637 No Yes No
    D1 84.135 1.175 No Yes No No No 0.602 No Yes No
    D2 85.241 1.245 No Yes No No Yes 0.437 No Yes No
    D3 85.477 1.382 No Yes No No Yes 0.210 No Yes No
    D4 73.096 1.587 No Yes No No No 0.560 No Yes No
    D5 85.174 1.097 No Yes No No No 0.441 No Yes No
    D6 83.283 1.168 No Yes No No No 0.557 No Yes No
    D7 93.770 0.783 No Yes No No Yes 0.417 No Yes No
    D8 93.715 0.794 No Yes No No Yes 0.364 No Yes No
    D9 94.821 0.871 No Yes No No No 0.319 No Yes No
    D10 95.057 1.005 No Yes No No Yes 0.274 No Yes No
    下载: 导出CSV
  • 加载中
WeChat 扫一扫看文章
计量
  • PDF下载量:  8
  • 文章访问数:  402
  • HTML全文浏览量:  66
文章相关
  • 发布日期:  2022-03-01
  • 收稿日期:  2021-05-27
  • 接受日期:  2021-07-27
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

/

返回文章

All Rights Reserved by the Chinese Chemical Society   中国化学会 版权所有 京ICP备05002797号

本系统由北京仁和汇智信息技术有限公司开发    技术支持: info@rhhz.net