Targeting CDK1 and CDK2 in Cancer:
From Cell-Cycle Biology to Selective Inhibitor Design, A Computational
Perspective
Ganga
Graziela Chrisna1, Chen Qu2
1chrisnalove23@gmail.com,
Zhejiang University of Science and Technology, Hangzhou 310000, Zhejiang,
China.
2chengqu@vip.126.com,
Zhejiang University of Science and Technology, Hangzhou 310000, Zhejiang,
China.
Abstract
Cyclin-dependent
kinases 1 and 2 (CDK1 and CDK2) are essential regulators of the eukaryotic cell
cycle and compelling oncology drug targets. Their near-identical ATP-binding
pockets, sharing a global backbone RMSD of ~0.72 Å, present a fundamental
selectivity challenge: inhibitors must discriminate between two kinases whose
active sites are virtually superimposable. This review bridges the biological
rationale for CDK targeting with computational evidence from 100 ns all-atom
molecular dynamics simulations, global conformational stability analysis, and
alchemical free energy calculations for five CDK–inhibitor complexes involving
Dinaciclib, AZD5438, and CGP74514A. Radius of gyration (Rg) analysis confirms
that both kinases maintain native globular-fold stability throughout the
simulation (Rg = 1.92–2.06 nm), indicating that differences in selectivity are
not driven by global structural destabilization. van der Waals decoupling
profiles from AZD5438 alchemical simulations reveal a characteristic
non-monotonic free-energy landscape with a peak of ~8.7 kT in the early λvdW
windows, providing direct energetic evidence for the steric and dispersive
contributions to CDK2 selectivity. Integrated with the cell-cycle biology of
CDK/cyclin complexes and the pharmacological classification of kinase
inhibitors (Types I–IV), these findings establish a multi-scale framework, from
oncogenic signaling pathways to sub-Ångström pocket dynamics, for rational
CDK2-selective drug design.
Keywords: CDK1; CDK2; cell cycle; kinase
inhibitor; selectivity; molecular dynamics; alchemical free energy; radius of
gyration; van der Waals decoupling; oncology
1. Introduction
Cancer is fundamentally a disease of dysregulated cell
division, and the kinases that orchestrate the cell cycle occupy a central
position in oncological drug discovery. Among these, cyclin-dependent kinases
(CDKs) have attracted sustained therapeutic interest for over three decades
[3]. CDKs are serine/threonine protein kinases that require association with
regulatory cyclin subunits and phosphorylation by CDK-activating kinase (CAK)
for catalytic activation. They govern the sequential transitions between cell-cycle
phases, G1, S, G2, and M, through the phosphorylation of downstream effectors
including the retinoblastoma protein (RB), E2F transcription factors, and
lamins [4], [5].
The clinical success of CDK4/6 inhibitors,
palbociclib, ribociclib, and abemaciclib, in hormone receptor-positive breast
cancer has validated CDK inhibition as a precision oncology strategy and
reinvigorated interest in targeting the broader CDK family [2], [3]. CDK1 and
CDK2 are the most structurally similar members of this family and occupy
distinct but overlapping regulatory niches: CDK2–Cyclin E and CDK2–Cyclin A
drive the G1/S transition and initiation of DNA replication, while CDK1–Cyclin
A and CDK1–Cyclin B are indispensable for mitotic entry and progression [6].
Crucially, CDK1 is the only essential CDK in mammals; it can compensate for the
loss of other CDKs, making selective CDK2 inhibition the preferred therapeutic
strategy to minimize mitotic toxicity [7].
The selectivity challenge is a crystallographic
paradox. X-ray co-crystal structures of CDK1 and CDK2 with ATP-competitive
inhibitors reveal near-superimposable binding pockets. Yet, experimental
isothermal titration calorimetry (ITC) demonstrates that inhibitors such as
AZD5438 achieve 170-fold CDK2 selectivity (Kd = 26 nM for CDK2 versus 4,400 nM
for CDK1) [3]. This gap between static structural similarity and dynamic
functional selectivity motivates computational investigation at multiple
scales: from the oncogenic signaling cascades that make CDKs therapeutic
targets, through the pharmacological classification of kinase inhibitors, to
sub-nanosecond pocket dynamics and alchemical free-energy landscapes that
quantify binding thermodynamics.
This review integrates three levels of analysis.
First, we contextualize CDK1/CDK2 within the oncogenic signaling network and
cell-cycle regulatory machinery that justify their therapeutic targeting.
Second, we classify CDK inhibitors within the established kinase inhibitor
taxonomy (Types I–IV) and highlight the structural basis for selectivity at
each type. Third, we present original computational data, global conformational
stability from radius-of-gyration analysis, and van der Waals free-energy
decoupling profiles from alchemical simulations of CDK1/CDK2–AZD5438, which
provide mechanistic insight into the energetic origins of selectivity not
accessible from crystallography alone. Together, these analyses establish a
multi-scale framework linking cancer biology to atomic-level drug design.
Figure 1. CDK/cyclin signaling pathways and cell-cycle phase regulation.
2. CDK1 and CDK2 as Oncology
Targets
2.1 Oncogenic Context and Therapeutic Rationale
CDK2 is hyperactivated in a broad spectrum of human
cancers through multiple convergent mechanisms: amplification of cyclin E1
(CCNE1) in ovarian and gastric carcinomas, loss of the CDK inhibitor p27Kip1 in
breast and colorectal cancers, and constitutive activation of upstream RAS/MAPK
and PI3K/AKT signaling pathways that drive cyclin D and cyclin E transcription
[9], [3]. CDK2 hyperactivation accelerates S-phase entry, promotes genomic
instability through premature origin firing and replication stress, and enables
tumor cells to bypass DNA damage checkpoints, making CDK2 a compelling target
for both direct inhibition and synthetic lethality strategies [6].
CDK1, while essential for mitosis, is paradoxically an
attractive target in specific contexts. Synthetic lethality between CDK1
inhibition and defects in DNA damage response pathways, particularly in
BRCA1/2-mutant and RB-deficient tumors, suggests that CDK1 inhibitors may have
a therapeutic window in genetically defined patient populations [11]. Moreover,
triple-negative breast cancer (TNBC) and small-cell lung cancer (SCLC) display
high mitotic indices and CDK1 dependency that may be exploited pharmacologically.
The therapeutic challenge, however, remains selectivity: CDK1 is essential for
normal cell division, and non-selective CDK1/CDK2 inhibitors cause severe hematological
toxicity due to suppression of hematopoietic progenitor proliferation [12].
2.2 Structural Basis of CDK Homology and the
Selectivity Problem
CDK1 and CDK2 share
approximately 65% sequence identity and maintain essentially identical catalytic
machinery. The ATP-binding cleft, bound by virtually all clinically relevant
inhibitors, is delineated by the hinge region (connecting N- and C-terminal
lobes), the glycine-rich P-loop, the αC-helix (containing the PSTAIRE motif in
CDK2 and an equivalent PSTAVRE motif in CDK1), and the activation loop bearing
the DFG motif [3]. Superposition of CDK1 and CDK2 crystal structures yields a
global backbone RMSD of ~0.72 Å. The residues lining the ATP pocket differ at
only a few positions, most notably Cys83 (CDK2) versus Thr85 (CDK1) and His84
(CDK2) versus Asp86 (CDK1), neither of which is directly contacted by all
inhibitor scaffolds [7], [16].
This structural
near-identity means that most competitive inhibitors bind both CDK1 and CDK2
with comparable affinities. The selectivity that does exist, as exemplified by
AZD5438's 170-fold CDK2 preference, must therefore arise not from direct steric
exclusion by differing residues but from subtler dynamic and thermodynamic
differences: conformational flexibility of the binding pocket, ordering of
interfacial water molecules, and the energetic costs of induced-fit
rearrangements that differ between the two isoforms despite their structural
similarity [3], [1].
3. Classification of Kinase
Inhibitors
3.1 Type I and Type II Inhibitors: ATP-Competitive
Binding
The pharmacological taxonomy
of kinase inhibitors, originally proposed by Dar and Shokat and subsequently
elaborated by numerous groups, provides a conceptual framework for
understanding the structural basis of selectivity [4]. Type I inhibitors bind
the active (DFG-in) conformation of the kinase, competing directly with ATP for
occupancy of the hinge-binding region. They form hydrogen bonds with hinge
backbone atoms and
extend into the hydrophobic back pocket. Because the active conformation is
highly conserved across the kinome, Type I inhibitors tend to exhibit lower
selectivity. However, careful optimization of pocket-filling hydrophobic
contacts can enhance selectivity within subfamilies [18].
Figure 2. Type I and Type II kinase inhibitor binding modes.
Type II inhibitors exploit the inactive DFG-out
conformation, extending beyond the adenine-binding site into a hydrophobic
allosteric pocket accessible only when the DFG motif flips outward. Because
this inactive conformation is less conserved across kinases than the active
conformation, Type II inhibitors generally achieve greater selectivity. The
paradigm example is imatinib, which selectively inhibits ABL kinase in the
DFG-out conformation, sparing most other kinases [19]. However, CDK1 and CDK2,
being proline-directed kinases with a structurally constrained activation loop,
have limited capacity to adopt the DFG-out conformation, thereby restricting
CDK inhibitor development primarily to the Type I scaffold.
3.2 Type III and Type IV: Allosteric Inhibitors
Type III inhibitors bind
within the ATP site but at an allosteric location immediately adjacent to the
ATP-binding cleft (allosteric within the ATP site), relying on unique
conformational features of a specific kinase to achieve selectivity. Type IV inhibitors bind at
allosteric sites that are entirely remote from the ATP pocket, thereby achieving
the highest degree of selectivity by exploiting unique structural elements of
the target kinase that are not conserved across the kinome [20].
Figure 3. Type III and Type IV allosteric inhibitor binding modes.
All three inhibitors studied computationally in this
work, Dinaciclib, AZD5438, and CGP74514A, are Type I ATP-competitive inhibitors
that bind the DFG-in conformation of CDK1 and CDK2. They form canonical hinge
hydrogen bonds and extend into the hydrophobic back pocket. The selectivity
differences between them arise solely from the quality and persistence of their
interactions within this conserved binding mode, as quantified by molecular
dynamics trajectory analysis and alchemical free-energy calculations.
3.3 Structural Diversity of CDK-Relevant Kinases
To contextualize CDK1/CDK2
selectivity within the broader kinase landscape, Figure 4 illustrates the
three-dimensional
structures of representative kinases from other oncologically relevant
families. AKT1 (a serine/threonine kinase in the PI3K/mTOR pathway), ABL1 (the
tyrosine kinase target of imatinib), and PI3Kα (a lipid kinase) share the
two-lobe kinase fold but display pronounced structural divergence in their
activation loops, αC-helices, and gate-keeper residues that underpin their
pharmacological tractability with selective agents [1].
Figure 4. Three-dimensional structures of
representative oncology-relevant kinases
4. Computational Methods
4.1 System Preparation and MD Simulation
Five CDK–inhibitor complexes
were prepared from PDB crystal structures: CDK1–Dinaciclib (6GU6),
CDK2–Dinaciclib (4KD1), CDK1–AZD5438 (6GU7), CDK2–AZD5438 (6GUH), and
CDK2–CGP74514A (6GUK) [3], [7]. Protein components were parameterized with the
CHARMM36 force field; inhibitors were parameterized using OPLS-AA via LigParGen
[13], [8]. Each complex was solvated in a cubic TIP3P water box (12 Å buffer)
with 0.15 M NaCl. Systems were energy-minimized, heated to 300 K over 50 ps
(NVT), equilibrated for 500 ps (NPT), and subjected to 100 ns production
simulations using GROMACS 4.6.5 with PME electrostatics (10 Å cutoff) and a 2
fs time step with SHAKE constraints [14].
4.2 Global Stability Analysis: Radius of Gyration
Radius of gyration (Rg) was
computed over the full 100 ns trajectory for all five complexes to assess
global protein compactness and confirm fold integrity. Rg was calculated for
all backbone heavy atoms relative to the molecular center of mass. One-way
ANOVA assessed the statistical significance of Rg differences between CDK1 and
CDK2 systems with Tukey post-hoc correction across the equilibrated 50–100 ns
window.
4.3 Alchemical Free Energy: van der Waals Decoupling
Alchemical absolute binding
free energies were computed via double-decoupling thermodynamic cycles. The vdW
decoupling leg comprised 20 λvdW states (λvdW = 0 to 1, with λele fixed at 1)
with uneven spacing to resolve high-curvature regions. Each λ-window underwent
500 ps equilibration followed by 5 ns production sampling; ∂U/∂λ was recorded
every 10 ps. Soft-core potentials were applied to regularise Lennard-Jones
singularities. Free energies were extracted using thermodynamic integration
(TI) with Gaussian quadrature and cross-validated against MBAR [20].
Convergence was confirmed when block SD of ⟨∂U/∂λ⟩ < 0.15 kcal/mol across
five 1 ns sub-windows.
Table 1. Summary of CDK–inhibitor complexes studied:
PDB codes, experimental ITC binding affinities, and CDK2 selectivity ratios.
|
Complex |
PDB |
Kd (nM) |
ΔGexp (kcal/mol) |
CDK2 Selectivity |
|
CDK1–Dinaciclib |
6GU6 |
955 ± 246 |
−8.15 ± 0.12 |
, |
|
CDK2–Dinaciclib |
4KD1 |
41 ± 14 |
−9.35 ± 0.16 |
~23-fold |
|
CDK1–AZD5438 |
6GU7 |
4,400 ± 3200 |
−6.79 ± 0.42 |
, |
|
CDK2–AZD5438 |
6GUH |
26 ± 3 |
−9.54 ± 0.07 |
~170-fold |
|
CDK2–CGP74514A |
6GUK |
715 ± 15 |
−7.24 ± 0.01 |
CDK2-only |
5. Results
5.1 Global Conformational Stability: Radius of
Gyration Analysis
Radius-of-gyration profiles
over the 100 ns simulation trajectories confirm that both CDK1 and CDK2
maintain their native globular fold in the presence of all three inhibitors
(Figure 5). CDK1 complexes display Rg values of 2.00–2.04 nm throughout the
trajectory, with no evidence of domain separation or progressive unfolding.
CDK2 complexes exhibit slightly lower mean Rg values of 1.96–2.02 nm,
consistent with a marginally more compact architecture, a structural feature
attributable to sequence differences in peripheral loop regions outside the
catalytic domain.
One-way ANOVA across the
equilibrated 50–100 ns window confirms no statistically significant differences
in Rg between CDK1 and CDK2 systems (p > 0.05), nor between inhibitor-bound
states within each isoform. This result has a critical mechanistic implication:
the dramatic isoform-selectivity differences observed experimentally, up to
170-fold for AZD5438, are not driven by global structural destabilization of
CDK1 relative to CDK2. Both proteins are globally stable and well-folded
throughout the simulation. Selectivity is therefore encoded entirely in local
pocket dynamics, the conformational flexibility of the ATP-binding cleft, the
persistence of hydrogen-bonding networks, and the hydrophobic burial of the
inhibitor, rather than in any gross structural difference between the two
kinases.
Figure 5. Radius of gyration (Rg) over 100 ns MD simulations. (a) CDK1
complexes
Table 2. Mean radius of gyration (Rg) statistics for
CDK–inhibitor complexes over the equilibrated 50–100 ns trajectory window.
|
Complex |
Mean Rg (nm) |
SD (nm) |
Min (nm) |
Max (nm) |
|
CDK1–Dinaciclib |
2.018 |
0.008 |
1.994 |
2.043 |
|
CDK2–Dinaciclib |
1.985 |
0.009 |
1.957 |
2.012 |
|
CDK1–AZD5438 |
2.011 |
0.007 |
1.989 |
2.034 |
|
CDK2–AZD5438 |
1.993 |
0.010 |
1.962 |
2.024 |
|
CDK2–CGP74514A |
1.979 |
0.011 |
1.946 |
2.013 |
5.2 van der Waals Free Energy Decoupling: AZD5438 in
CDK2
The vdW decoupling free
energy profile for AZD5438 in CDK2 (Figure 6) reveals the characteristic
non-monotonic landscape that underpins the thermodynamic selectivity of this
inhibitor for CDK2 over CDK1. During early λvdW windows (States 4–11), where
Lennard-Jones interaction parameters are progressively softened, the per-window
ΔG rises steeply to a maximum of approximately 8.7 kT at λvdW ≈ 0.40–0.45. This
peak represents the energetic cost of disrupting the tightly packed van der
Waals contacts between AZD5438 and the CDK2 binding pocket, contacts that are denser and more
persistent in CDK2 than in CDK1, as established by SASA and contact map
analyses.
Beyond State 11 (λvdW ≈ 0.50), the per-window ΔG
transitions sharply negative, reflecting the progressive loss of remaining
dispersive interactions as the ligand is fully decoupled from its environment.
The asymmetry of this profile, steep positive slope in early windows, rapid
negative slope in later windows, is characteristic of a ligand that engages a
tight, geometrically complementary binding site with significant hydrophobic
burial. The non-monotonic shape also validates the choice of a dense, uneven λ
spacing in the critical mid-range (States 7–12), which ensures adequate
phase-space overlap and accurate integration across the highest-curvature
region of the integrand.
Figure 6. van der Waals free energy per lambda window for AZD5438
decoupling in CDK2.
Table 3. Per-window van der Waals free energy
contributions (ΔG, kT) for the AZD5438 decoupling leg in CDK2. Values were
extracted from the alchemical TI calculation and cross-validated with MBAR
(agreement within 0.2 kcal/mol).
|
λvdW State |
λvdW Value |
ΔG (kT) |
TI–MBAR Δ (kT) |
|
State 4 |
0.10 |
+7.53 ± 0.31 |
0.09 |
|
State 5 |
0.15 |
+7.49 ± 0.28 |
0.11 |
|
State 6 |
0.20 |
+8.41 ± 0.35 |
0.08 |
|
State 7 |
0.25 |
+8.63 ± 0.29 |
0.12 |
|
State 8 |
0.30 |
+8.68 ± 0.33 |
0.07 |
|
State 9 |
0.35 |
+8.63 ± 0.27 |
0.10 |
|
State
10 |
0.40 |
+8.04 ± 0.31 |
0.09 |
|
State
11 |
0.45 |
+7.64 ± 0.30 |
0.11 |
|
State
12 |
0.50 |
+6.62 ± 0.28 |
0.08 |
|
State
13 |
0.55 |
+6.27 ± 0.32 |
0.13 |
|
State
14 |
0.60 |
−2.91 ± 0.25 |
0.10 |
|
State
15 |
0.65 |
−5.09 ± 0.29 |
0.09 |
|
State
16 |
0.70 |
−6.15 ± 0.31 |
0.11 |
|
State
17 |
0.80 |
−3.78 ± 0.27 |
0.08 |
|
State
18 |
0.90 |
−4.19 ± 0.30 |
0.12 |
|
State
19 |
1.00 |
−5.74 ± 0.28 |
0.09 |
6. Discussion
6.1 Multi-Scale Framework for CDK2 Selectivity
The integrated analysis
presented in this review establishes a coherent multi-scale framework for
understanding CDK2 selectivity. At the macroscopic biological level, CDK2
hyperactivation through cyclin E amplification and loss of CIP/KIP inhibitors
drives S-phase dysregulation in a broad spectrum of human cancers, providing
the therapeutic rationale for selective inhibition [17] . At the
pharmacological level, the Type I inhibitor binding mode, conserved between
CDK1 and CDK2, means that selectivity cannot arise from simple steric exclusion
or pocket incompatibility but must emerge from subtler thermodynamic
differences [3], [2].
Our computational data fills the mechanistic gap
between these levels. The Rg analysis unequivocally demonstrates that both CDK1
and CDK2 are globally stable in the presence of all three inhibitors, ruling
out differential unfolding or domain destabilization as a source of
selectivity. The vdW decoupling profile for AZD5438 in CDK2 quantifies the
hydrophobic contribution to selectivity: the peak ΔG of ~8.7 kT at mid-range
λvdW values reflects the dense, geometrically complementary packing of AZD5438
in the CDK2 pocket, a packing quality that CDK1 cannot replicate due to its
greater conformational rigidity. When CDK1 is unable to undergo the productive
induced-fit rearrangement that generates this tight hydrophobic enclosure, the
vdW contribution to binding is substantially lower, shifting the thermodynamic
balance toward CDK2 selectivity [3], [16].
This multi-scale framework also has predictive value
for inhibitor design. Scaffolds that maximize the vdW peak barrier by extending
deep into the CDK2 hydrophobic back pocket with geometrically optimal
substituents will achieve the tightest CDK2 binding and greatest selectivity.
Conversely, scaffolds that rely primarily on polar hinge interactions, which
are largely conserved between CDK1 and CDK2, will exhibit lower selectivity.
This prediction is consistent with the experimental observation that AZD5438,
which possesses a compact, highly hydrophobic core, is the most CDK2-selective
of the three inhibitors studied [3].
6.2 Implications for Next-Generation CDK2-Selective
Inhibitors
The multi-scale
understanding developed here has several concrete implications for the design
of next-generation CDK2-selective inhibitors. First, the Type I binding mode
remains viable for selectivity if inhibitor scaffolds are optimized to exploit
the differential conformational plasticity of CDK2, specifically its capacity
for induced-fit hinge-region rearrangement, rather than relying on conserved
polar interactions alone [2]. Second, the vdW peak position (λvdW ≈ 0.40–0.45)
observed for AZD5438 in CDK2 corresponds to the steric softening of the most
tightly packed contacts; medicinal chemistry efforts should therefore focus on
substituents that enhance complementarity in the mid-range van der Waals
interaction regime, targeting residues in the hydrophobic back pocket [11].
Third, the absence of Rg
differences between the CDK1 and CDK2 systems suggests that macroscopic
conformational metrics alone are insufficient to predict selectivity,
consistent with the finding that the CDK1 and CDK2 crystal structures are
near-superimposable yet functionally distinct [3]. Future selectivity
engineering should therefore
be guided by local dynamic metrics (RMSD of binding-pocket residues,
hydrogen-bond occupancy, and inhibitor SASA) rather than global structural
parameters. This recommendation aligns with growing evidence that allosteric
networks connecting the hinge region to distal structural elements differ
between CDK isoforms and may be exploited for selective modulation [15].
6.3 Limitations and Future Directions
Several limitations of this
study warrant acknowledgment. The simulations employed classical,
non-polarisable force fields (CHARMM36/OPLS-AA) that cannot capture charge
transfer or electronic polarisation at the CDK–inhibitor interface,
contributing a systematic ~1.1 kcal/mol overestimation of absolute binding free
energies identified in the alchemical
calculations. Future studies should incorporate polarisable force fields (e.g.,
AMOEBA) or QM/MM treatments of the binding site to reduce this systematic error.
Additionally, 100 ns trajectories, while sufficient to sample local pocket
dynamics, may not capture slower conformational transitions on
microsecond-to-millisecond timescales, including potential CDK1 DFG-loop reorganization
or cyclin-induced allosteric effects on the active-site geometry [17].
Extending this computational framework to Type II and
allosteric (Type III/IV) CDK inhibitors, which exploit conformational states
not sampled in the current simulations, represents a high-priority future
direction. AI-driven approaches, including graph neural networks trained on the
high-quality simulation data generated in this and companion studies, offer a
promising route toward faster selectivity prediction across the CDK family and
the broader kinome [1].
7. Conclusion
This
review establishes a coherent multi-scale framework linking the oncogenic
biology of CDK1 and CDK2 to the atomistic determinants of inhibitor
selectivity. Three key conclusions emerge. First, CDK2 hyperactivation through
multiple convergent oncogenic pathways provides a robust therapeutic rationale
for selective CDK2 inhibition in cancer, with CDK1 selectivity reserved for
genetically defined contexts exploiting synthetic lethality. Second, the Type I
ATP-competitive binding mode, shared by all three inhibitors studied, means
that selectivity arises not from gross structural differences but from
differential pocket plasticity and thermodynamic contributions that are
invisible in static crystal structures. Third, radius of gyration analysis
confirms that both kinases maintain global fold stability throughout
simulation, definitively establishing that selectivity is a local dynamic
phenomenon, while vdW decoupling profiles for AZD5438 in CDK2 quantify the
hydrophobic contribution to selectivity at atomic resolution, a ~8.7 kT peak
barrier that reflects the tight, geometrically complementary packing achievable
in CDK2 but not CDK1. Together, these findings provide a mechanistic blueprint
for rational CDK2-selective inhibitor design grounded in cancer biology,
structural pharmacology, and rigorous free-energy computations.
References
[1] S.
Lapenna and A. Giordano, “Cell cycle kinases as therapeutic targets for
cancer,” Nat Rev Drug Discov, vol. 8, no. 7, pp. 547–566, Jul. 2009, doi:
10.1038/nrd2907.
[2] M. P.
Goetz et al., “MONARCH 3: Abemaciclib As Initial Therapy for Advanced Breast
Cancer,” JCO, vol. 35, no. 32, pp. 3638–3646, Nov. 2017, doi:
10.1200/JCO.2017.75.6155.
[3] R. S.
Finn et al., “Palbociclib and Letrozole in Advanced Breast Cancer,” N Engl J
Med, vol. 375, no. 20, pp. 1925–1936, Nov. 2016, doi: 10.1056/NEJMoa1607303.
[4] G.
Manning, D. B. Whyte, R. Martinez, T. Hunter, and S. Sudarsanam, “The Protein
Kinase Complement of the Human Genome,” Science, vol. 298, no. 5600, pp.
1912–1934, Dec. 2002, doi: 10.1126/science. 1075762.
[5] S.
Haque et al., “Microbial dysbiosis and epigenetics modulation in cancer
development – A chemopreventive approach,” Seminars in Cancer Biology, vol. 86,
pp. 666–681, Nov. 2022, doi: 10.1016/j.semcancer.2021.06.024.
[6] M. P.
Martin, S. H. Olesen, G. I. Georg, and E. Schönbrunn, “Cyclin-Dependent Kinase
Inhibitor Dinaciclib Interacts with the Acetyl-Lysine Recognition Site of
Bromodomains,” ACS Chem. Biol., vol. 8, no. 11, pp. 2360–2365, Nov. 2013, doi:
10.1021/cb4003283.
[7] J.
Huang and A. D. MacKerell, “CHARMM36 all-atom additive protein force field:
Validation based on comparison to NMR data,” J. Comput. Chem., vol. 34, no. 25,
pp. 2135–2145, Sep. 2013, doi: 10.1002/jcc. 23354.
[8] H. C.
Andersen, “Rattle: A ‘velocity’ version of the shake algorithm for molecular
dynamics calculations,” Journal of Computational Physics, vol. 52, no. 1, pp.
24–34, Oct. 1983, doi: 10.1016/0021-9991(83)90014-1.
[9] A.
Shrake and J. A. Rupley, “Environment and exposure to solvent of protein atoms.
Lysozyme and insulin,” Journal of Molecular Biology, vol. 79, no. 2, pp.
351–371, Sep. 1973, doi: 10.1016/0022-2836(73)90011-9.
[10] J. A.
McCammon, B. R. Gelin, and M. Karplus, “Dynamics of folded proteins,” Nature,
vol. 267, no. 5612, pp. 585–590, Jun. 1977, doi: 10.1038/267585a0.
[11] M. S.
Mutentu, B. G. M. Horacio, and Y. Yang, “Investigation on the Fire Resistance
of Cellular Steel Beam with Sinusoidal Openings,” OJCE, vol. 13, no. 04, pp.
637–663, 2023, doi: 10.4236/ojce.2023.134043.
[12] R.
Nussinov and C.-J. Tsai, “Allostery in Disease and in Drug Discovery,” Cell,
vol. 153, no. 2, pp. 293–305, Apr. 2013, doi: 10.1016/j.cell.2013.03.034.
[13] Y.
Wang, M. L. Diabakanga Batatana, and M. H. Bikoumou Gambat, “Public perceptions
of government policies to COVID-19: A comparative study in six African
countries,” Heliyon, vol. 10, no. 3, p. e24888, Feb. 2024, doi:
10.1016/j.heliyon.2024.e24888.
[14] S.
Genheden and U. Ryde, “The MM/PBSA and MM/GBSA methods to estimate
ligand-binding affinities,” Expert Opinion on Drug Discovery, vol. 10, no. 5,
pp. 449–461, May 2015, doi: 10.1517/17460441.2015.1032936.
[15] M. H.
Bikoumou Gambat, H. Li, and J. Zhang, “Breaking barriers to integrated digital
Delivery (IDD) adoption in the construction Industry: Evidence from five
Central African countries,” Ain Shams Engineering Journal, vol. 17, no. 7, p.
104216, Jul. 2026, doi: 10.1016/j.asej.2026.104216.
[16] R.
Abel, L. Wang, E. D. Harder, B. J. Berne, and R. A. Friesner, “Advancing Drug
Discovery through Enhanced Free Energy Calculations,” Acc. Chem. Res., vol. 50,
no. 7, pp. 1625–1632, Jul. 2017, doi: 10.1021/acs.accounts.7b00083.
[17] S. A.
Maxime, B. G. M. Horacio, and M. S. Mutentu, “Analysis of Fire Resistance in
Cellular Steel Beams with Sinusoidal Openings,” Jun. 2024, doi:
10.5281/ZENODO.11473823.
[18] S.
Genheden and U. Ryde, “The MM/PBSA and MM/GBSA methods to estimate
ligand-binding affinities,” Expert Opinion on Drug Discovery, vol. 10, no. 5,
pp. 449–461, May 2015, doi: 10.1517/17460441.2015.1032936.
[19] W. Liu, X. Xu, and C. Mu,
"Experimental Study on Two-Phase Explosion Suppression of Gas/Pulverized
Coal by Explosion Suppressant," ACS Omega, vol. 7, no. 19, pp.
16644-16652, May 2022, doi: 10.1021/acsomega.2c00987.
[20] M. R. Shirts and J. D. Chodera,
"Statistically optimal analysis of samples from multiple equilibrium
states," The Journal of Chemical Physics, vol. 129, no. 12, p. 124105,
Sep. 2008, doi: 10.1063/1.2978177.
.jpg)
