Activated Cdc42-associated tyrosine kinase 1 (ACK1) is a member of the VIII tyrosine kinase family (Manser et al., 1993; Lin et al., 2012). The overexpression of ACK1 has a potential impact on other types of diseases also, but recently research studies are mainly focused on its driving effect on cancers (Wang A. et al., 2021). ACK1 can integrate signals of various receptor tyrosine kinases (RTKs), can transfer extracellular signals from receptor tyrosine kinases to cytoplasmic and nuclear effectors, and in turn may regulate the expression levels of some RTKs (Mahajan et al., 2010; Mahajan and Mahajan 2010; Fox et al., 2019) Previous studies indicated that ACK1 regulates the activity of androgen receptor (AR) by tyrosine phosphorylation to increase the growth of hormone-refractory prostate cancers (Mahajan et al., 2005; Mahajan et al., 2012; Mahajan and Mahajan, 2013; Mahajan and Mahajan, 2015). Next-generation sequencing (NGS) studies suggested that recurrent ACK1 gene amplification and somatic mutations in different types of cancers lead to neoplastic transformation. ACK1 is an epigenetic regulator (Wang A. et al., 2021), and its coding gene TNK2 is related to hematological malignancies and other types of cancers. Since epigenetics is a reversible process, targeting ACK1 could reverse the epigenetic changes and the existence of drug resistance in malignancies. Recently, various approaches, such as fragment-based drug design, repurposing, skeleton transition, and high-throughput screening, have been used to design, discover, and synthesize several highly efficient and specific inhibitors to target ACK1. Reports indicated that 71 FDA-approved small-molecule kinase inhibitors (SMKIs) and additional 16 SMKIs are approved by other authorities. The clinical trials of SMKIs showed that approximately 110 novel kinases are currently being explored as targets. Many SMKIs are in clinical development for diseases other than cancers (Attwood et al., 2021).

The physiological role of ACK1 is involved in human brains (La Torre et al., 2013), inflammation (Zhao et al., 2018), and the immune system. To date, the ACK1 gene is seen in 131 missense mutations, 39 nonsense mutations, and three fusion mutations within domains of ACK1 in 21 types of cancers (Prieto-Echague et al., 2010; Maxson et al., 2016; Wang Z.-Z. et al., 2021). The latest trends in the inhibitor design include allosteric and covalent inhibitors, bifunctional inhibitors, and chemical degraders (Attwood et al., 2021). FDA has approved many inhibitors which have good therapeutic effects and are used for multiple targets. These multikinase inhibitors can provide potential strategies for simultaneously targeting ACK1 and other targets.

In this study, 14 multikinase inhibitors, namely, GNF-7 (1), (2), (3), (4), dasatinib (5), bosutinib (6), ceritinib (7), PD158780 (8), vemurafenib (9), ADZ9291 (10), sunitinib (11), flavopiridol (12), and gefitinib (13), (14) are selected (Wang A. et al., 2021). Some of these inhibitors have entered clinical trials, and others are approved by FDA as therapeutics for other targets. GNK-7 (1) is used to double-target ACK1 for the NRAS-mutated acute myeloid leukemia cells (Choi et al., 2010; Cho et al., 2018). Complex (2) has a better impact on Ba/F3-NRAS-G12D cells, while Complex (3) exhibits good pharmacokinetic properties. Complex (4) shows inhibitory activities toward Ba/F3-NRAS-G12D and OCI-acute myeloid leukemia (AML3) stability (Wang Z.-Z. et al., 2021). The reference Complex dasatinib (5) (Lombardo et al., 2004) and Complex bosutinib (6) (Stansfield et al., 2013) are Abelson leukemia virus (ABL) and proto-oncogene tyrosine-protein kinase SRC (SRC) kinase inhibitors. Ceritinib (7) is an anaplastic lymphoma kinase (ALK) fusion protein inhibitor (Verduzco et al., 2018). PD158780 (8) is an adenosine 5′-triphosphate (ATP) competitive epidermal growth factor receptor (EGFR) and ACK1 inhibitor (Nur-E-Kamal et al., 2005), while vemurafenib (9) is a potent inhibitor targeting BRAF (V600E). Zelboraf (10) showed selectivity toward several non-rapidly accelerated fibrosarcoma (RAF) kinases (Bollag et al., 2010). Sunitinib (11), flavopiridol (12), and gefitinib (13) are identified as ACK1 inhibitors (Phatak and Zhang, 2013). Complex (14) showed inhibitory activity toward 26 kinases. (Wang A. et al., 2021). As experimental studies and clinical trials are expensive and time-consuming, computational approaches help to predict the chemical and biological properties of complexes and their binding affinities to targets in a better manner. Dasatinib (BMS-354825 or Sprycel) and AIM-100 (Liu et al., 2010; Mahajan et al., 2014) were initially assessed as inhibitors for ACK1 signaling in vitro and in vivo. Due to its multitarget activity against many tyrosine kinases, dasatinib (5) is selected as a reference inhibitor in this work. The chemical structures of these 14 multikinase inhibitors are given in Supplementary Figure 1. The main objective of this work was to find out which inhibitors would have the best characteristics as a drug candidate.

The density functional theory has been an effective method to predict drug–target interactions in recent years. The conceptual density functional theory (CDFT), developed by Prof. R.G. Parr et al. (Parr and Yang 1989; von Szentpaly 2000; von Szentpaly 2017; von Szentpaly 2018a; von Szentpaly 2018b; von Szentpaly 2020; Kohn et al., 1996; Ayers and Yang 2003; Parr and Yang 1995; Sarkar and Chattaraj 2021; Chattaraj et al., 2002; Chattaraj et al., 2003; Parr et al., 1999), has been used to study the molecular descriptors. The validity, physical basis, and limitations of these descriptors (Kaya and Kaya 2015a; Kaya and Kaya 2015b; Von Szentpály et al., 2020) are being investigated in recent years, and they provide deep insights into the reactivity of the complexes. Few parameters from CDFT-based computational peptidology (CDFT-CP) developed by Prof. Glossmann Mitnik et al (Frau et al., 2018; Flores-Holguín et al., 2019a; Flores-Holguín et al., 2019b; Frau et al., 2019; Flores-Holguín et al., 2020a; Flores-Holguín et al. 2020b; Flores-Holguín et al. 2020c; Flores-Holguín et al. 2021) has been incorporated to analyze the physicochemical parameters of these multikinase inhibitors. Previous results related to global descriptors have shown useful predictions for new drug entities (NCEs) (Srivastava, 2021a) and also for drug-like small complexes (Srivastava, 2021b).

The structures of inhibitors have been optimized with M06-2X (Zhao and Truhlar, 2008)/6-31G (d,p) basis sets from G16 software suites (Frisch et al., 2016) with the PCM (Marenich et al., 2009) solvent model. Different conformational structures are optimized, and then the structure with lower minima is selected for further studies. Vibrational frequency analysis is carried out to confirm that there is no negative frequency for the lower minima structures. The optimized structures of the multikinase inhibitors are given in Figure 1. GaussView is used to draw the optimized structures for the inhibitors (Dennington et al., 2003). The TDDFT calculations are carried out on the ground state optimized structures with M06-2X/6-31G (d,p) basis sets using water as a solvent. GaussSum is used to visualize the absorption wavelength, oscillatory strength, and the number of transitions (O’boyle et al., 2008).

The output (pdb) structures are used to generate the SMILES notation for the inhibitors from the online SMILES translator and structure file generator tool, which is given in Table 1.

Various molecular properties and bioactivity scores of inhibitors are calculated by Molinspiration Chemoinformatics tools (https://www.molinspiration.com/). (accessed, December 2021) The pharmacokinetic parameters such as mutagenic effect, irritant effect, tumorigenicity, and effect on the reproductive system are estimated by Osiris Property Explorer, 2021 (www.organicchemistry.org/prog/peo/). Other toxicity parameters are predicted by BOILED-Egg in Swiss ADME (Diana and Zoete 2016). Simulations Plus ADMET Predictor (Simulations Plus Inc, 2022) is used to predict the physicochemical and biopharmaceutical (PCB), Salmonella typhimurium reverse mutation assay (AMES) mutagenicity, toxicity, and risk prediction. The online SwissTargetPrediction tool is used to predict efficient protein targets for these drug complexes (Diana et al., 2019).

The structure of ACK1 (pdb id-6VQM) is optimized with the MMF4 force field using Avogadro software (Hanwell et al., 2016), as shown in Figure 2A. Various conformers have been taken for ACK1 structures, and MD simulations are carried out for the lower conformational structure using CABS-flex 2.0 (Kuriata et al., 2018) coarse-grained protein modeling tools. The protein near-native dynamics were obtained from 10 nanosecond MD simulations (all-atom, explicit water, and for all protein metafolds with force fields). Figure 2B shows the best MD-simulated ACK1 model with the root-mean-square fluctuation (RMSF) graph. RMSF measures the average deviation of protein residues over time from the reference values. The PrankWeb server (Jendele et al., 2019) is used to identify potential binding pockets (Pitsillou et al., 2020) for ACK1, which is shown in Figure 2C.

Docking studies for ACK1–inhibitors are carried out by Autodock 4.2.6 packages (Morris et al., 2009). The empirical free energy function and the Lamarckian genetic algorithm were used with other default parameters. Among the top ten binding poses, the model with lower interacting binding energies was selected to study protein–ligand binding interactions by protein–ligand binding profiler (PLIP) online tools (Salentin et al., 2015).

DFT-based molecular descriptors are numerical characterizations of structural features of complexes. The description of the preferred sites provides a firm explanation for the reactivity of the complexes. The formulas used to calculate the global descriptors are given in Supplementary Information. Global reactivity descriptors and highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) values are given in Table 2. The ionization potential (IP) values of all studied complexes are larger, which reflects the ability to lose electrons more easily. All complexes have IP values (4.39–7.62 eV) except Complex 14 (4.11 eV). Electron affinity (EA) reflects the ability of an atom to gain electrons which is lower (0.06–2.4 eV) for most of the complexes. The chemical hardness correlates to the stability, and the reactivity of these inhibitors is measured by the softness. The higher electrophilicity index represents the more reactive nature of these inhibitors, i.e., (1.70–5.53) eV. HOMO and LUMO values of complexes are predictive measures of their interaction with the target complexes. The higher HOMO energy reflected more reactive molecules in the reactions with electrophiles, while lower LUMO energy is necessary for molecular reactions with nucleophiles. The smaller HOMO–LUMO energy gap for these inhibitors corresponds to better stability. Comparing the hardness and HOMO–LUMO gap results, it can be anticipated that hardness can be considered an exact measure, while HOMO–LUMO gap can be taken as an approximation for the inhibitors.

The wavelength (nm), oscillatory strength, and number of transitions for these multikinase inhibitors are given in Table 3. The TDDFT results showed that the absorption spectra of these complexes lie in the region (246–361 nm), except for Complex 4, which has an absorption wavelength in the visible region (541.35 nm). The percentage for HOMO→LUMO transitions is higher for all inhibitors, which indicates that the interaction between HOMO→ LUMO is enough to lower the energy of the state below one of the HOMO→LUMO states. All DFT results have positive correlations toward the corresponding results of the reference complex dasatinib (5).

The molecular parameters such as clog P, topological polar surface area (TPSA), hydrogen bond donors (HBD), hydrogen bond acceptors (HBA), and molecular weight (MW) are given in Table 4. For orally active drugs, Lipinski’s ‘‘rule of five” states that 1) molecular weight (MW) < 500; 2) the calculated octanol/water partition coefficient (clogP) < 5; 3) there were fewer than five hydrogen bond donors (HBD) (OH and NH groups); and 4) there are less than ten hydrogen bond acceptors (HBA) (notably N and O) (Lipinski et al., 1997). clogP are >5 for complexes 4, 9, and 14, while complexes 2 and 4 have HBA values >10. All the complexes have HBD <5. For the drug-like hits based on the Muegge (Bayer) criteria (Muegge et al., 2001) (200 ≤ MW ≤ 600, −2 ≤ LogP ≤ 5, TPSA ≤ 150, HBD ≤ 5, HBA ≤ 10, and RotB ≤ 15), only complexes 2 and 4 have violation for HBA > 10, so they cannot be considered drug-like hits. It is known that more violations of the Lipinski rules lead to bioavailability problems. The drug-likeliness, drug score, mutagenicity, tumorigenicity, irritant effect, and reproductive effect are given in Table 5. The result indicates that Complex 5 has high toxicity for irritant and reproductive effects. Complex 6 has high toxicity for mutagenicity, tumorigenicity, and reproductive effect, while Complex 7 shows high irritant effect. Complexes 9 and 10 have higher mutagenic effect, while Complex 11 has higher irritant effect. Complex 10 has mild toxicity effect for tumorigenicity, irritant effect, and reproductive effect.

The Biopharmaceutics Classification System (BCS) has defined five criteria for the permeability of drugs: 1) absolute bioavailability or mass balance studies in humans, 2) urinary recovery of unchanged drug in humans, 3) in vivo intestinal perfusion studies in humans, 4) in vitro permeation studies across a monolayer of cultured epithelial cells, and/or 5) high metabolism as defined under Biopharmaceutical Drug Disposition Classification System (BDDCS) (Ku 2008; Chen et al., 2011). Few of these parameters are given in Table 6, which are predicted by BOILED-Egg in Swiss ADME. Most of the studied complexes showed an inhibitory effect for multitargets. All the inhibitors have Log K _p values that lie in the range of (−6.0 to 7.0). No BBB permeability is predicted for all inhibitors except Complex 8. The GI absorption is higher for complexes 3, 5, 6, 8, and 10–13.

The PCB, toxicity, and risk prediction from Simulations Plus ADMET Predictor shows varied PCM, ADMET, and risk prediction for all the inhibitors, as shown in Figure 3. The transport and metabolic properties of these inhibitors are given in Supplementary Figure 4 and Supplementary Figure 5, respectively. All complexes showed good inhibitory activities, but variation in absorption, distribution, metabolism, excretion, toxicity (ADMET), biopharmaceutical properties, and risk prediction indicate that these inhibitors may be used for combination therapies.

SwissTargetPrediction results indicated that these complexes can be used as multikinase inhibitors. Complexes 1–13 have kinase as the main potential target (blue color), while complexes 10, 11, 12, and 14 have the potential to target protease, transferase, enzyme, family A G protein-coupled receptor, and voltage-gated ion channel, as shown in Figure 4. The multitargeting activities of these inhibitors are validated by the experimental results (Wang Z.-Z. et al., 2021).

The docking of ACK1–inhibitors along with their binding energies is given in Supplementary Figure 2. It has been observed that drug-like and lead-like criteria for physiochemical properties were only applied to ligands whose receptor BE < −6 kcal/mol for the best docking pose. All the studied complexes have the best binding energy poses (BE < −6 kcal/mol), which reflect the importance of these non-receptors as potential druggable targets for target therapies. The binding of inhibitors is mostly toward the best binding pocket of ACK1, as predicted in Figure 3C.

Protein–Ligand Interaction Profiler (PLIP) predicted hydrophobic interactions (HYs) and hydrogen bonds (HB) in all ACK1–inhibitors, which enhance the binding affinity and biological activity of complex molecules and help in stabilizing the biochemical environments. Other interactions such as salt bridge (SB) was observed in ACK1-(3–6, 10, 13) complexes. SB and π–π stacking (perpendicular) interactions were observed in (ACK1-1, 9) complexes. π–π stacking (parallel) and no SB were seen in ACK1–9 complexes. ACK1–12 complexes have cation–π and SB interactions, while only cation–π interactions are seen in ACK1–10 complexes, as shown in Supplementary Figure 3. The π–π stacking is essential for the favorable electron correlation, whereas cation–π contacts produce further electrostatic contributions (Brylinski, 2018). These noncovalent interactions promote high-affinity binding of inhibitors to their targets which is important for the success of rational drug discovery.

DFT results showed that the studied 14 inhibitors are highly stable. Most of the inhibitors are non-mutagenic, non-tumorigenic, non-irritant, and without effects on reproduction, and these inhibitors have good drug score values. Since cancer is a heterogeneous disease and variations have been observed in toxicity and ADMET results, it is recommended to use these inhibitors as combination drugs instead of single-drug treatment. ACK1–multikinase inhibitors show better binding affinities. They can be used as potential drugs for treating cancers and other diseases. Due to the multitarget activities of these inhibitors, we hope that these studies will also provide a reference and possibility for the study and treatment of other diseases in the future (Rahman et al., 2021).