Applying deep learning to iterative screening of medium-sized molecules for protein-protein interaction-targeted drug discovery.

We combined a library of medium-sized molecules with iterative screening using multiple machine learning algorithms that were ligand-based, which resulted in a large increase of the hit rate against a protein-protein interaction target. This was demonstrated by inhibition assays using a PPI target, Kelch-like ECH-associated protein 1/nuclear factor erythroid 2-related factor 2 (Keap1/Nrf2), and a deep neural network model based on the first-round assay data showed a highest hit rate of 27.3%. Using the models, we identified novel active and non-flat compounds far from public datasets, expanding the chemical space.

DLiP-PPI library: An integrated chemical database of small-to-medium-sized molecules targeting protein-protein interactions.

Protein-protein interactions (PPIs) are recognized as important targets in drug discovery. The characteristics of molecules that inhibit PPIs differ from those of small-molecule compounds. We developed a novel chemical library database system (DLiP) to design PPI inhibitors. A total of 32,647 PPI-related compounds are registered in the DLiP. It contains 15,214 newly synthesized compounds, with molecular weight ranging from 450 to 650, and 17,433 active and inactive compounds registered by extracting and integrating known compound data related to 105 PPI targets from public databases and published literature. Our analysis revealed that the compounds in this database contain unique chemical structures and have physicochemical properties suitable for binding to the protein-protein interface. In addition, advanced functions have been integrated with the web interface, which allows users to search for potential PPI inhibitor compounds based on types of protein-protein interfaces, filter results by drug-likeness indicators important for PPI targeting such as rule-of-4, and display known active and inactive compounds for each PPI target. The DLiP aids the search for new candidate molecules for PPI drug discovery and is available online (https://skb-insilico.com/dlip).

Identification of novel inhibitors of Keap1/Nrf2 by a promising method combining protein-protein interaction-oriented library and machine learning.

Protein-protein interactions (PPIs) are prospective but challenging targets for drug discovery, because screening using traditional small-molecule libraries often fails to identify hits. Recently, we developed a PPI-oriented library comprising 12,593 small-to-medium-sized newly synthesized molecules. This study validates a promising combined method using PPI-oriented library and ligand-based virtual screening (LBVS) to discover novel PPI inhibitory compounds for Kelch-like ECH-associated protein 1 (Keap1) and nuclear factor erythroid 2-related factor 2 (Nrf2). We performed LBVS with two random forest models against our PPI library and the following time-resolved fluorescence resonance energy transfer (TR-FRET) assays of 620 compounds identified 15 specific hit compounds. The high hit rates for the entire PPI library (estimated 0.56-1.3%) and the LBVS (maximum 5.4%) compared to a conventional screening library showed the utility of the library and the efficiency of LBVS. All the hit compounds possessed novel structures with Tanimoto similarity ≤ 0.26 to known Keap1/Nrf2 inhibitors and aqueous solubility (AlogP < 5). Reasonable binding modes were predicted using 3D alignment of five hit compounds and a Keap1/Nrf2 peptide crystal structure. Our results represent a new, efficient method combining the PPI library and LBVS to identify novel PPI inhibitory ligands with expanded chemical space.

A combination of 19F NMR and surface plasmon resonance for site-specific hit selection and validation of fragment molecules that bind to the ATP-binding site of a kinase.

19F NMR has recently emerged as an efficient, sensitive tool for analyzing protein binding to small molecules, and surface plasmon resonance (SPR) is also a popular tool for this purpose. Herein a combination of 19F NMR and SPR was used to find novel binders to the ATP-binding pocket of MAP kinase extracellular regulated kinase 2 (ERK2) by fragment screening with an original fluorinated-fragment library. The 19F NMR screening yielded a high primary hit rate of binders to the ERK2 ATP-binding pocket compared with the rate for the SPR screening. Hit compounds were evaluated and categorized according to their ability to bind to different binding sites in the ATP-binding pocket. The binding manner was characterized by using isothermal titration calorimetry and docking simulation. Combining 19F NMR with other biophysical methods allows the identification of multiple types of hit compounds, thereby increasing opportunities for drug design using preferred fragments.

Development of novel CXC chemokine receptor 7 (CXCR7) ligands: selectivity switch from CXCR4 antagonists with a cyclic pentapeptide scaffold.

The CXC chemokine receptor 7 (CXCR7)/ACKR3 is a chemokine receptor that recognizes stromal cell-derived factor 1 (SDF-1)/CXCL12 and interferon-inducible T-cell α chemoattractant (I-TAC)/CXCL11. Here, we report the development of novel CXCR7-selective ligands with a cyclic pentapeptide scaffold through an SAR study of CXC chemokine receptor 4 (CXCR4) selective antagonist FC131 [cyclo(-d-Tyr-l-Arg-l-Arg-l-Nal-Gly-), Nal = 3-(2-naphthyl)alanine]. Substitution of Gly with l-Pro switched the receptor preference of the peptides from CXCR4 to CXCR7. The SAR study led to the identification of a potent CXCR7 ligand, FC313 [cyclo(-d-Tyr-l-Arg-l-MeArg-l-Nal-l-Pro-)], which recruits ß-arrestin to CXCR7. Investigations via receptor mutagenesis and molecular modeling experiments suggest a possible binding mode of the cyclic pentapeptide CXCR7 agonist.

Identification of a Dual Inhibitor of SRPK1 and CK2 That Attenuates Pathological Angiogenesis of Macular Degeneration in Mice.

Excessive angiogenesis contributes to numerous diseases, including cancer and blinding retinopathy. Antibodies against vascular endothelial growth factor (VEGF) have been approved and are widely used in clinical treatment. Our previous studies using SRPIN340, a small molecule inhibitor of SRPK1 (serine-arginine protein kinase 1), demonstrated that SRPK1 is a potential target for the development of antiangiogenic drugs. In this study, we solved the structure of SRPK1 bound to SRPIN340 by X-ray crystallography. Using pharmacophore docking models followed by in vitro kinase assays, we screened a large-scale chemical library, and thus identified a new inhibitor of SRPK1. This inhibitor, SRPIN803, prevented VEGF production more effectively than SRPIN340 owing to the dual inhibition of SRPK1 and CK2 (casein kinase 2). In a mouse model of age-related macular degeneration, topical administration of eye ointment containing SRPIN803 significantly inhibited choroidal neovascularization, suggesting a clinical potential of SRPIN803 as a topical ointment for ocular neovascularization. Thus SRPIN803 merits further investigation as a novel inhibitor of VEGF.

Mutation analysis and molecular modeling for the investigation of ligand-binding modes of GPR84.

GPR84 is a G protein-coupled receptor for medium-chain fatty acids. Capric acid and 3,3'-diindolylmethane are specific agonists for GPR84. We built a homology model of a GPR84-capric acid complex to investigate the ligand-binding mode using the crystal structure of human active-state ß2-adrenergic receptor. We performed site-directed mutagenesis to subject ligand-binding sites to our model using GPR84-Giα fusion proteins and a [(35)S]GTPγS-binding assay. We compared the activity of the wild type and mutated forms of GPR84 by [(35)S]GTPγS binding to capric acid and diindolylmethane. The mutations L100D `Ballesteros-Weinstein numbering: 3.32), F101Y (3.33) and N104Q (3.36) in the transmembrane helix III and N357D (7.39) in the transmembrane helix VII resulted in reduced capric acid activity but maintained the diindolylmethane responses. Y186F (5.46) and Y186H (5.46) mutations had no characteristic effect on capric acid but with diindolylmethane they significantly affected the G protein activation efficiency. The L100D (3.32) mutant responded to decylamine, a fatty amine, instead of a natural agonist, the fatty acid capric acid, suggesting that we have identified a mutated G protein-coupled receptor-artificial ligand pairing. Our molecular model provides an explanation for these results and interactions between GPR84 and capric acid. Further, from the results of a double stimulation assay, we concluded that diindolylmethane was a positive allosteric modulator for GPR84.

Antidiabetic sulfonylureas and cAMP cooperatively activate Epac2A.

Sulfonylureas are widely used drugs for treating insulin deficiency in patients with type 2 diabetes. Sulfonylureas bind to the regulatory subunit of the pancreatic ß cell potassium channel that controls insulin secretion. Sulfonylureas also bind to and activate Epac2A, a member of the Epac family of cyclic adenosine monophosphate (cAMP)-binding proteins that promote insulin secretion through activation of the Ras-like guanosine triphosphatase Rap1. Using molecular docking simulation, we identified amino acid residues in one of two cyclic nucleotide-binding domains, cNBD-A, in Epac2A predicted to mediate the interaction with sulfonylureas. We confirmed the importance of the identified residues by site-directed mutagenesis and analysis of the response of the mutants to sulfonylureas using two assays: changes in fluorescence resonance energy transfer (FRET) of an Epac2A-FRET biosensor and direct sulfonylurea-binding experiments. These residues were also required for the sulfonylurea-dependent Rap1 activation by Epac2A. Binding of sulfonylureas to Epac2A depended on the concentration of cAMP and the structures of the drugs. Sulfonylureas and cAMP cooperatively activated Epac2A through binding to cNBD-A and cNBD-B, respectively. Our data suggest that sulfonylureas stabilize Epac2A in its open, active state and provide insight for the development of drugs that target Epac2A.

Optimized method of G-protein-coupled receptor homology modeling: its application to the discovery of novel CXCR7 ligands.

Homology modeling of G-protein-coupled seven-transmembrane receptors (GPCRs) remains a challenge despite the increasing number of released GPCR crystal structures. This challenge can be attributed to the low sequence identity and structural diversity of the ligand-binding pocket of GPCRs. We have developed an optimized GPCR structure modeling method based on multiple GPCR crystal structures. This method was designed to be applicable to distantly related receptors of known structural templates. CXC chemokine receptor (CXCR7) is a potential drug target for cancer chemotherapy. Homology modeling, docking, and virtual screening for CXCR7 were carried out using our method. The predicted docking poses of the known antagonists were different from the crystal structure of human CXCR4 with the small-molecule antagonist IT1t. Furthermore, 21 novel CXCR7 ligands with IC50 values of 1.29-11.4 µM with various scaffolds were identified by structure-based virtual screening.

Effect of the STAT3 inhibitor STX-0119 on the proliferation of cancer stem-like cells derived from recurrent glioblastoma.

Signal transducer and activator of transcription (STAT) 3, a member of a family of DNA-binding molecules, is a potential target in the treatment of cancer. The highly phosphorylated STAT3 in cancer cells contributes to numerous physiological and oncogenic signaling pathways. Furthermore, a significant association between STAT3 signaling and glioblastoma multiforme stem-like cell (GBM-SC) development and maintenance has been demonstrated in recent studies. Previously, we reported a novel small molecule inhibitor of STAT3 dimerization, STX-0119, as a cancer therapeutic. In the present study, we focused on cancer stem-like cells derived from recurrent GBM patients and investigated the efficacy of STX-0119. Three GBM stem cell lines showed many stem cell markers such as CD133, EGFR, Nanog, Olig2, nestin and Yamanaka factors (c-myc, KLF4, Oct3/4 and SOX2) compared with parental cell lines. These cell lines also formed tumors in vivo and had similar histological to surgically resected tumors. STAT3 phosphorylation was activated more in the GBM-SC lines than serum-derived GB cell lines. The growth inhibitory effect of STX-0119 on GBM-SCs was moderate (IC50 15-44 µM) and stronger compared to that of WP1066 in two cell lines. On the other hand, the effect of temozolomide was weak in all the cell lines (IC50 53-226 µM). Notably, STX-0119 demonstrated strong inhibition of the expression of STAT3 target genes (c-myc, survivin, cyclin D1, HIF-1α and VEGF) and stem cell-associated genes (CD44, Nanog, nestin and CD133) as well as the induction of apoptosis in one stem-like cell line. Interestingly, VEGFR2 mRNA was also remarkably inhibited by STX-0119. In a model using transplantable stem-like cell lines in vivo GB-SCC010 and 026, STX-0119 inhibited the growth of GBM-SCs at 80 mg/kg. STX-0119, an inhibitor of STAT3, may serve as a novel therapeutic compound against GBM-SCs even in temozolomide-resistant GBM patients and has the potential for GBM-SC-specific therapeutics in combination with temozolomide plus radiation therapy.

Dr. PIAS 2.0: an update of a database of predicted druggable protein-protein interactions.

Druggable Protein-protein Interaction Assessment System (Dr. PIAS) is a database of druggable protein-protein interactions (PPIs) predicted by our support vector machine (SVM)-based method. Since the first publication of this database, Dr. PIAS has been updated to version 2.0. PPI data have been increased considerably, from 71,500 to 83,324 entries. As the new positive instances in our method, 4 PPIs and 10 tertiary structures have been added. This addition increases the prediction accuracy of our SVM classifier in comparison with the previous classifier, despite the number of added PPIs and structures is small. We have introduced the novel concept of 'similar positives' of druggable PPIs, which will help researchers discover small compounds that can inhibit predicted druggable PPIs. Dr. PIAS will aid the effective search for druggable PPIs from a mine of interactome data being rapidly accumulated. Dr. PIAS 2.0 is available at http://www.drpias.net.

Structure-activity relationship study of a CXC chemokine receptor type 4 antagonist, FC131, using a series of alkene dipeptide isosteres.

A structure-activity relationship study on a highly potent CXC chemokine receptor type 4 (CXCR4) antagonist, FC131 [cyclo(-d-Tyr(1)-Arg(2)-Arg(3)-Nal(4)-Gly(5)-)], was carried out using a series of alkene isosteres of the d-Tyr(1)-l/d-Arg(2) dipeptide to investigate the binding mode of FC131 and its derivatives with CXCR4. The structure-activity relationships of isostere-containing FC131 analogues were similar to those of the parent FC131 and its derivatives, suggesting that a trans-conformer of the d-Tyr(1)-Arg(2) peptide bond is the dominant contributor to the bioactive conformations of FC131. Although NMR analysis demonstrated that the two conformations of the peptidomimetic containing the d-Tyr(1)-d-Arg(2) isostere are possible, binding-mode prediction indicated that the orientations of the alkene motif within d-Tyr(1)-MeArg(2) peptidomimetics depend on the chirality of Arg(2) and the ß-methyl group of the isostere unit, which makes the dominant contribution for binding to the receptor. The most potent FC122 [cyclo(-d-Tyr(1)-d-MeArg(2)-Arg(3)-Nal(4)-Gly(5)-)] bound with CXCR4 by a binding mode different from that of FC131.

Molecular modeling study of cyclic pentapeptide CXCR4 antagonists: new insight into CXCR4-FC131 interactions.

CXCR4 is a G-protein coupled receptor that is associated with many diseases such as breast cancer metastasis, HIV infection, leukemic disease and rheumatoid arthritis, and is thus considered an attractive drug target. Previously, we identified a cyclic pentapeptide, FC131, that is a potent antagonist for CXCR4. In this study, we constructed a three dimensional model of the CXCR4-FC131 complex. To investigate the backbone flexibility of FC131, we performed molecular dynamics simulations of FC131 based on the NMR structure of FC131, and obtained snapshot structures from the trajectories which were used to model the docking pose of FC131 into CXCR4. Our final model of the CXCR4-FC131 complex is partially different from the X-ray crystal structure of CXCR4-CVX15 and suggests water-mediated interactions. Nevertheless, this docking pose is consistent with the experimental data. We believe our model will aid in the discovery and development of small-molecule antagonists for CXCR4.

Virtual screening and further development of novel ALK inhibitors.

Anaplastic lymphoma kinase (ALK) has been in the spotlight in recent years as a promising new target for therapy of non-small-cell lung cancer (NSCLC). Since the identification of the echinoderm microtubule-associated protein-like 4 (EML4)-ALK fusion gene in some NSCLC patients was reported in 2007, various research groups have been seeking ALK inhibitors. Above all, crizotinib (PF-02341066) has been under clinical trial, and its therapeutic efficacy of inhibiting ALK in NSCLC has been reported. Among anticancer drugs, drug resistance appears frequently necessitating various kinds of inhibitors. We identified novel ALK inhibitors by virtual screening from the public chemical library collected by the Chemical Biology Research Initiative (CBRI) at the University of Tokyo, and inhibitors that are more potent were developed.

Antitumor activity of a novel small molecule STAT3 inhibitor against a human lymphoma cell line with high STAT3 activation.

Signal transducer and activator of transcription (STAT)3, a member of a family of DNA-binding molecules mediating numerous physiological and oncogenic signaling pathways, is a novel target in cancer cells which show high phosphorylation of STAT3. Recently, we identified a novel small-molecule inhibitor of STAT3 dimerization, STX-0119, as a cancer therapeutic. We investigated the mechanisms responsible for the antitumor activity in vitro and in vivo through numerous biochemical and biological assays. Specifically, the effects of STX-0119 on target genes (c-myc, cyclin D1, survivin) and apoptosis induction were analyzed in tumors treated with STX-0119 in vivo. STX-0119 showed strong growth-inhibitory activity against a broad range of hematological cancer cell lines, particularly lymphomas. STX-0119 suppressed the growth of SCC3 cells, a human lymphoma cell line with highly activated STAT3, through apoptosis and down-regulation of STAT3 targets such as c-myc, cyclin D1, survivin and Bcl-xL. Notably, Tyr-705-phosphorylated STAT3 up-regulation was not significantly suppressed by STX-0119, as opposed to other STAT3 inhibitors. STX-0119 demonstrated potent antitumor effects in vivo in SCC3-bearing nude mice by way of the down-regulation of STAT3 target genes and induction of apoptosis in the tumors. Thus, STX-0119 may be a new type of STAT3 inhibitor exhibiting strong antitumor activity.

Dr. PIAS: an integrative system for assessing the druggability of protein-protein interactions.

BACKGROUND: The amount of data on protein-protein interactions (PPIs) available in public databases and in the literature has rapidly expanded in recent years. PPI data can provide useful information for researchers in pharmacology and medicine as well as those in interactome studies. There is urgent need for a novel methodology or software allowing the efficient utilization of PPI data in pharmacology and medicine. RESULTS: To address this need, we have developed the 'Druggable Protein-protein Interaction Assessment System' (Dr. PIAS). Dr. PIAS has a meta-database that stores various types of information (tertiary structures, drugs/chemicals, and biological functions associated with PPIs) retrieved from public sources. By integrating this information, Dr. PIAS assesses whether a PPI is druggable as a target for small chemical ligands by using a supervised machine-learning method, support vector machine (SVM). Dr. PIAS holds not only known druggable PPIs but also all PPIs of human, mouse, rat, and human immunodeficiency virus (HIV) proteins identified to date. CONCLUSIONS: The design concept of Dr. PIAS is distinct from other published PPI databases in that it focuses on selecting the PPIs most likely to make good drug targets, rather than merely collecting PPI data.

Identification of novel ASK1 inhibitors using virtual screening.

Apoptosis signal-regulating kinase 1 (ASK1, also called MAP3K5) is a mitogen-activated protein kinase kinase kinase (MAP3K) that plays important roles in stress-induced cell death and inflammation, and is expected as a new therapeutic target for cancer, cardiovascular diseases, and neurodegenerative diseases. We identified novel ASK1 inhibitors by virtual screening from the public chemical library collected by Chemical Biology Research Initiative (CBRI) at the University of Tokyo.

Evaluation of docking calculations on X-ray structures using CONSENSUS-DOCK.

We are participating in the challenge of identifying active compounds for target proteins using structure-based virtual screening (SBVS). We use an in-house customized docking program, CONSENSUS-DOCK, which is a customized version of the DOCK4 program in which three scoring functions (DOCK4, FlexX and PMF) and consensus scoring have been implemented. This paper compares the docking calculation results obtained using CONSENSUS-DOCK and DOCK4, and demonstrates that CONSENSUS-DOCK produces better results than DOCK4 for major X-ray structures obtained from the Protein Data Bank (PDB).

Structure-activity relationship of novel DAPK inhibitors identified by structure-based virtual screening.

Death-associated protein kinase (DAPK) is a serine/threonine protein kinase implicated in diverse programmed cell death pathways. DAPK is a promising target protein for the treatment of ischemic diseases. We identified novel potent and selective DAPK inhibitors efficiently by structure-based virtual screening, then further developed the hit compounds. In this paper, we describe the development of the hit compounds and the structure-activity relationship studies of the DAPK inhibitors in detail, including calculation of the solvated interaction energy (SIE), and verification of selectivity using a kinase panel.

[In-silico approaches for fragment-based drug design].

An important step to promote fragment-based drug design (FBDD) is to find high-quality fragment molecules. Therefore the design of the fragment library is the most crucial stage. In our fragment library, the main considerations are ligand efficiency (LE), diversity, and solubility with drug-like properties. We especially considered LE to raise hit probability in screening. We estimated LE of the fragment molecule based on known LE values of the active compounds. We also developed a docking program suitable for screening fragments rather than drug compounds. Furthermore, we explored fragment-linking program, which links together fragments that bind to adjacent sites on a target protein so as to promote FBDD in silico.