Compounds with immunomodulatory activity and their therapeutic uses
Through the energy map neural network model combined with local and global characteristics, an efficient PD-1/PD-L1 inhibitor was developed, which solved the problem of low efficiency in the development of small and medium-sized drugs in the existing technology, and achieved efficient inhibitor identification and synthesis.
Patent Information
- Application Number
- CN202180033571.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-03-11
- Filing Date
- 2021-03-05
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-03-05
AI Technical Summary
The lack of efficient small molecule drugs during the development of existing PD-1/PD-L1 inhibitors, traditional methods are time-consuming and labor-intensive and difficult to quickly identify new skeletons inefficiently, and existing machine learning models exist in predicting drug-target interactions insufficient.
An energy graph neural network (EGNN) model was developed to combine local and global features to select and synthesize new PD-1/PD-L1 inhibitors through bootstrap machine learning methods, leveraging topological structures and binding structures within the bag Interactions are superior to traditional methods in predicting the efficacy of inhibitors.
A highly effective PD-1/PD-L1 inhibitor with an IC50 value of 339.9 nM was successfully identified and synthesized, which was superior to known biologically active compounds, and demonstrated excellent performance on the structural diversity dataset, significantly improving the drug Design efficiency and accuracy.
Smart Images

Figure CN115551497B_ABST
Abstract
Description
[0001] Government support clause
[0002] This invention was made with government support under CA023168 awarded by the National Institutes of Health. The U.S. Government has certain rights in this invention.
[0003] CROSS REFERENCE TO RELATED APPLICATIONS
[0004] This patent application is related to and claims the benefit of priority of U.S. Provisional Application Serial No. 62 / 987,914, filed on March 11, 2020, the contents of which are hereby incorporated by reference in their entirety. Technical Field
[0005] The present invention generally relates to compounds having immunomodulatory activity and their therapeutic uses. Also described herein are pharmaceutical compositions of such compounds and methods of treating cancer patients by administering a therapeutically effective amount of such compounds alone, together with other therapeutic agents or in pharmaceutical compositions.
[0006] background
[0007] This section introduces aspects that may help facilitate a better understanding of the present disclosure. Therefore, these statements should be read in this light and should not be construed as admissions of what is or is not prior art.
[0008] Programmed cell death protein 1 (PD-1) is an immune checkpoint receptor implicated in the creation of new cancer therapeutics 1 Long-term interaction between the T-cell receptor and the major histocompatibility complex (MHC) leads to upregulation of PD-1 on the surface of activated T-cells 2 Activated T cells produce cytokines, such as interferon-γ, which in turn cause tumor cells to express programmed death ligand 1 (PD-L1) on their cell surface. 2 Tumors evade the immune system by exploiting the interaction between PD-1 and its ligand PD-L1, resulting in reduced effector T cell function and survival, leading to a suppressive immune response in the tumor microenvironment. 2 Inhibition of PD-1 / PD-L1 interaction can enhance anti-tumor immunity, and a great deal of work has been done to develop monoclonal antibodies as inhibitors of PD-1 / PD-L1 interaction. 3,4 For example, pembrolizumab, cemiplimab, and nivolumab are three FDA-approved anti-PD-1 antibodies 4 The discovery of small molecule inhibitors will have advantages over antibodies, such as rapid onset of action, simple in vivo administration, and the ability to penetrate cell membranes and interact with the cytoplasmic domains of cell surface receptors.5 Over the past few years, significant progress has been made in designing PD-1 / PD-L1 inhibitors. 6,7 Specifically, Bristol-Myers Squibb (BMS) discovered a group of potent PD-1 / PD-L1 small molecule inhibitors based on peptide mimetic molecules and non-peptide small molecules. 6,7 In particular, BMS revealed a chemical library containing a 2-methyl-3-biphenyl-methanol backbone. Later, Holak et al. studied the interaction of BMS molecules with PD-L1, showing that BMS molecules induce PD-L1 dimerization, and also reported the crystal structure of a compound with dimeric PD-L1. 8,9 Based on these findings, we envisioned developing a machine learning (ML) framework for selecting and testing new PD-1 / PD-L1 inhibitors. There is an unmet medical need for more effective cancer treatments, especially PD-1 / PD-L1 inhibitors with immunomodulatory activity. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] The above and other objects, features and advantages of the present invention will become more apparent when taken in conjunction with the following description and accompanying drawings, in which:
[0011] Figure 1 .The EGNN model utilizes a combination of local (A) and global (B) features. Local features are computed from the molecular graph of a molecule using a GNN to assign weights to each subgraph of the molecule. Global features are a collection of docking scores used to represent the interaction between a compound and a protein. These two features are combined to create a concatenated vector (C), which is passed through a SoftMax layer and bootstrapped to classify molecules as having "low" or "high" potency for PD-1 / PD-L1 interaction.
[0012] Figure 2A , Top: Classification of training data in BMS and Incyte patents; Bottom left: Main PD-L1 inhibitor skeleton in BMS patent. R group can be CN, Cl, Br or CH3; and Bottom right: Main PD-L1 inhibitor skeleton in Incyte patent. Here A and B represent sub-skeletons.
[0013] Figure 2B and 2C Heatmaps of pairwise Tanimoto similarity scores are shown for BMS and Incyte compounds, respectively.
[0014] Figures 3A-3C. The light pink chain represents the PD-1 protein, and the light cyan chain represents the PD-L1 protein in the PD-1 / PD-L1 complex crystal structure (PDB ID: 4ZQK). The wheat-colored chain represents the PD-L1 chain A, and the bluish-white chain represents the PD-L1 chain B in the PD-L1 homodimer crystal structure (PDB ID: 5N2F). ( Figure 3A ) Overlapping and aligned crystal structures of PD-1 / PD-L1 (4ZQK) and PD-L1 dimer (5N2F). ( Figure 3B ) Two crystal structures superimposed and aligned, with a defined binding site (grey grid) for the PD-L1 dimer (5N2F). Figure 3C ) Crystal structure of PD-L1 dimer (5N2F) with a small molecule (ligand ID: 8HW) in its binding site (grey grid).
[0015] Figure 4A shows the training-validation and testing scheme used for the model;
[0016] Figure 4B Cohen’s κ scores of EGNN and GNN with different training-validation and test sets are depicted;
[0017] Figure 4C The F1 scores of EGNN and GNN models with different training-validation and test sets are shown; and Figure 4D Heatmap showing pairwise Tanimoto similarity scores between precision-recall curves for BMS and Incyte compounds for EGNN, GNN, RF, and SVM models trained using Incyte data.
[0018] Figure 5A It shows that EGNN predicted a new PD-1 / PD-L1 inhibitor compound 4b through skeleton hopping of BMS compound 4a or BMS-1 and BMS-1002. The blue part of 4b is added from BMS-1002, and the pink part is added from 4a (BMS-1).
[0019] Figure 5B The position of compound 4b in the top docking pose in the PD-L1 homodimer crystal structure (PDB ID: 5N2F) is shown. The inset shows the hydrophobic tunnel of compound 4b.
[0020] Figure 5C The chemical interactions of compound 4b in the top docking pose in the PD-L1 homodimer are shown. The blue and pink parts show the bonds (sticks) of 4b. A Thr20 and A The yellow lines between Ala121 and Tyrosine represent hydrogen bonds.A The orientation of the aromatic ring of Tyr56 suggests a plausible π-π interaction with the blue aromatic ring of 2,3-dihydro-1,4-benzodioxine in compound 4b.
[0021] Figure 5D The IC values of 4a (BMS-1 control compound, red) and new compound 4b (blue) are shown. 50 DMSO controls for the assayed positive (PC-DMSO, purple) and negative controls (NC-DMSO, green) are shown for each assay concentration.
[0022] Detailed Description
[0023] For the purpose of promoting an understanding of the principles of the present disclosure, reference will now be made to the embodiments illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the present disclosure is intended thereby.
[0024] As used herein, the following terms and phrases shall have the following meanings. Unless defined otherwise, all technical and scientific terms used herein have the same meanings as commonly understood by one of ordinary skill in the art.
[0025] In the present disclosure, the term "about" can allow for a certain degree of variability in values or ranges, for example, within 10%, within 5%, or within 1% of a stated value or a stated range limit. In the present disclosure, the term "substantially" can allow for a certain degree of variability in values or ranges, for example, within 90%, within 95%, or within 99% of a stated value or a stated range limit.
[0026] In this document, unless the context clearly indicates otherwise, the terms "a", "an" or "the" are used to include one or more than one. Unless otherwise indicated, the term "or" is used to refer to a non-exclusive "or". In addition, it should be understood that the words or terms used herein, and in the absence of additional definition, are used only for descriptive purposes and not for limiting purposes. Any use of section headings is intended to assist the reading of the document and should not be interpreted as limiting. In addition, information related to section headings may appear within or outside of that particular section. In addition, all publications, patents, and patent documents mentioned in this document are incorporated herein by reference as a whole, as if individually incorporated by reference. If there is an inconsistency between the usage of this document and those documents so incorporated by reference, the usage in the incorporated references should be considered to supplement the usage of this document; for irreconcilable inconsistencies, the usage in this document shall prevail.
[0027] "Halogen" refers to F, CI, Br or I. "Halogen substitution" or "halo" substitution refers to the replacement of one or more hydrogen atoms with F, CI, Br or I.
[0028] The term "alkyl" as used herein refers to a saturated monovalent chain of carbon atoms, which may be optionally branched. It should be understood that in embodiments comprising alkyl, exemplary variants of those embodiments include lower alkyl groups, such as C1-C6 alkyl, methyl, ethyl, propyl, 3-methylpentyl, etc.
[0029] The term "alkenyl" as used herein refers to an unsaturated monovalent chain of carbon atoms including at least one double bond, which may be optionally branched. It should be understood that in embodiments including alkenyl, exemplary variants of those embodiments include lower alkenyl, such as C2-C6, C2-C4 alkenyl, etc.
[0030] The term "alkynyl" as used herein refers to an unsaturated monovalent chain of carbon atoms including at least one triple bond, which may be optionally branched. It should be understood that in embodiments including alkynyl, exemplary variants of those embodiments include lower alkynyls, such as C2-C6, C2-C4 alkynyls, etc.
[0031] The term "cycloalkyl" as used herein refers to a monovalent chain of carbon atoms, a portion of which forms a ring. It should be understood that in embodiments comprising cycloalkyl, exemplary variants of those embodiments include lower cycloalkyls, such as C3-C8 cycloalkyl, cyclopropyl, cyclohexyl, 3-ethylcyclopentyl, etc.
[0032] The term "cycloalkenyl" used herein refers to an unsaturated monovalent chain of carbon atoms, a portion of which forms a ring. It should be understood that in embodiments comprising cycloalkenyl, exemplary variants of those embodiments include lower cycloalkenyls, such as C3-C8, C3-C6 cycloalkenyls.
[0033] The term "alkylene" as used herein refers to a saturated divalent chain of carbon atoms, which may be optionally branched. It should be understood that in embodiments comprising alkylene, exemplary variants of those embodiments include lower alkylene, such as C2-C4 alkylene, methylene, ethylene, propylene, 3-methylpentylene, etc.
[0034] It should be understood that each of the alkyl, cycloalkyl, alkenyl, cycloalkenyl, alkylene and heterocycle can be optionally substituted with independently selected groups such as alkyl, haloalkyl, hydroxyalkyl, aminoalkyl, carboxylic acids and derivatives thereof including esters, amides and nitrites, hydroxy, alkoxy, acyloxy, amino, alkyl and dialkylamino, acylamino, thio, and the like, and combinations thereof.
[0035] The term "heterocyclic" or "heterocycle" as used herein refers to a monovalent chain of carbon and heteroatoms, wherein the heteroatoms are selected from nitrogen, oxygen and sulfur, and a portion thereof, at least one heteroatom, forms a ring. The term "heterocycle" may include both "aromatic heterocycle" and "non-aromatic heterocycle". Heterocycles include 4-7 membered monocyclic and 8-12 membered bicyclic rings, such as imidazolyl, thiazolyl, oxazolyl, oxazinyl, thiazinyl, dithianyl, dioxanyl, isoxazolyl, isothiazolyl, triazolyl, furanyl, tetrahydrofuranyl, dihydrofuranyl, pyranyl, tetrazolyl, pyrazolyl, pyrazinyl, pyridazinyl, imidazolyl, pyridinyl, pyrrolyl, dihydropyrrolyl, pyrrolidinyl, piperidinyl, piperazinyl, pyrimidinyl, morpholinyl, tetrahydrothienyl, thienyl, azetidinyl, oxetanyl, thiirane, oxiranyl, aziridine, etc. "Heterocycle" may be optionally substituted at any one or more positions capable of carrying a hydrogen atom.
[0036] The term "aryl" used herein includes monocyclic and polycyclic aromatic carbocyclic groups, each of which can be optionally substituted. The term "optionally substituted aryl" refers to an aromatic monocyclic or polycyclic ring of a carbon atom, such as phenyl, naphthyl, etc., which can be optionally substituted by one or more independently selected substituents, such as halogen, hydroxyl, amino, alkyl or alkoxy, alkylsulfonyl, cyano, nitro, etc.
[0037] The term "heteroaryl" or "aromatic heterocycle" includes substituted or unsubstituted aromatic monocyclic structures, preferably 5- to 7-membered rings, more preferably 5- to 6-membered rings, whose ring structures include at least one heteroatom, preferably one to four heteroatoms, more preferably one or two heteroatoms. The term "heteroaryl" may also include ring systems having one or two rings, wherein at least one ring is heteroaromatic, for example, the other cyclic ring may be a cycloalkyl, cycloalkenyl, cycloalkynyl, aromatic carbocycle, heteroaryl and / or heterocycle. Heteroaryl includes, but is not limited to, pyridyl, N-oxopyridyl, pyrimidinyl, pyrazinyl, pyridazinyl, triazinyl, furanyl, quinolyl, isoquinolyl, thienyl, imidazolyl, thiazolyl, indolyl, pyrrolyl, oxazolyl, benzofuranyl, benzothienyl, benzothiazolyl, isoxazolyl, pyrazolyl, triazolyl, tetrazolyl, indazolyl, 1,2,4-thiadiazolyl, isothiazolyl, benzothienyl, purinyl, carbazolyl, benzimidazolyl, indolyl, etc. In certain embodiments, the heteroaryl group has from 1 to about 20 carbon atoms, and in other embodiments, from about 3 to about 20 carbon atoms. In certain embodiments, the heteroaryl group contains 3 to about 14, 3 to about 7, or 5 to 6 ring atoms. In certain embodiments, the heteroaryl group has 1 to about 4, 1 to about 3, or 1 to 2 heteroatoms.
[0038] In certain embodiments, "heterocycloalkyl" refers to a non-aromatic heterocycle, wherein one or more of the ring-forming atoms are heteroatoms such as O, N or S atoms.Heterocycloalkyl groups can include monocyclic or polycyclic (for example, with 2,3 or 4 fused rings) ring systems and spirocycles.Example heterocycloalkyl groups include morpholino, thiomorpholino, piperazinyl, tetrahydrofuranyl, tetrahydrothienyl, 2,3-dihydrobenzofuranyl, 1,3-benzodioxole, benzo -1,4-dioxane, piperidinyl, pyrrolidinyl, isoxazolidinyl, isothiazolidinyl, pyrazolidinyl, oxazolidinyl, thiazolidinyl, imidazolidinyl etc. Also included in the definition of heterocycloalkyl are parts with one or more aromatic rings fused (i.e., with a common bond) to non-aromatic heterocycles, such as heterocyclic phthalimido, naphthalimido and benzo derivatives.Heterocycloalkyl groups with one or more fused aromatic rings can be connected by aromatic or non-aromatic moieties.
[0039] As used herein, the term "optionally substituted" or "optionally substituted" means that the group in question is unsubstituted or substituted with one or more of the specified substituents. When the group in question is substituted with more than one substituent, the substituents may be the same or different. In addition, when the terms "independently," "independently are," and "independently selected from" are used, it is meant that the groups in question may be the same or different. Certain terms defined herein may appear more than once in a structure, and after such appearance, each term will be defined independently of the others.
[0040] The term "patient" includes humans and non-human animals such as companion animals (dogs and cats, etc.) and livestock animals. Livestock animals are animals raised for food production. The patient to be treated is preferably a mammal, particularly a human.
[0041] The term "pharmaceutically acceptable carrier" is recognized in the art and refers to a pharmaceutically acceptable material, composition or vehicle involved in carrying or transporting any subject composition or its components, such as a liquid or solid filler, diluent, excipient, solvent or encapsulating material. Each carrier must be "acceptable" in the sense of being compatible with the subject composition and its components and not injurious to the patient. Some examples of materials that can serve as pharmaceutically acceptable carriers include: (1) sugars, such as lactose, glucose and sucrose; (2) starches, such as corn starch and potato starch; (3) cellulose and its derivatives, such as sodium carboxymethyl cellulose, ethyl cellulose and cellulose acetate; (4) Powdered tragacanth; (5) malt; (6) gelatin; (7) talc; (8) excipients such as cocoa butter and suppository waxes; (9) oils such as peanut oil, cottonseed oil, safflower oil, sesame oil, olive oil, corn oil and soybean oil; (10) glycols such as propylene glycol; (11) polyols such as glycerol, sorbitol, mannitol and polyethylene glycol; (12) esters such as ethyl oleate and ethyl laurate; (13) agar; (14) buffers such as magnesium hydroxide and aluminum hydroxide; (15) alginic acid; (16) pyrogen-free water; (17) isotonic saline; (18) Ringer's solution; (19) ethanol; (20) phosphate buffered saline; and (21) other nontoxic compatible substances employed in pharmaceutical formulations.
[0042] The term "administering" as used herein includes all ways of introducing the compounds and compositions described herein into a patient, including, but not limited to, oral (po), intravenous (iv), intramuscular (im), subcutaneous (sc), transdermal, inhalation, buccal, ocular, sublingual, vaginal, rectal, etc. The compounds and compositions described herein can be administered in unit dosage forms and / or formulations containing conventional non-toxic pharmaceutically acceptable carriers, adjuvants and vehicles.
[0043] Solid pharmaceutical forms may contain inert ingredients and carrier substances such as calcium carbonate, calcium phosphate, sodium phosphate, lactose, starch, mannitol, alginates, gelatin, guar gum, magnesium stearate, aluminum stearate, methylcellulose, talc, highly dispersed silicic acid, silicone oil, higher molecular weight fatty acids (such as stearic acid), gelatin, agar or vegetable or animal fats and oils or solid high molecular weight polymers (such as polyethylene glycol); if desired, preparations suitable for oral administration may contain additional flavoring and / or sweeteners.
[0044] Liquid drug forms can be sterilized, and / or comprise auxiliary substances in appropriate cases, such as preservatives, stabilizers, wetting agents, penetrants, emulsifiers, spreading agents, solubilizers, salts, sugars or sugar alcohols for regulating osmotic pressure or for buffering, and / or viscosity modifiers. The example of such additives is tartrate and citrate buffers, ethanol and masking agents (such as ethylenediaminetetraacetic acid and non-toxic salts thereof). High molecular weight polymers such as liquid polyethylene oxide, microcrystalline cellulose, carboxymethyl cellulose, polyvinyl pyrrolidone, dextran or gelatin are suitable for regulating viscosity. The example of solid carrier material is starch, lactose, mannitol, methylcellulose, talcum powder, highly dispersed silicic acid, high molecular weight fatty acids (such as stearic acid), gelatin, agar, calcium phosphate, magnesium stearate, animal and vegetable fats and solid high molecular weight polymers, such as polyethylene glycol.
[0045] Oily suspensions for parenteral or topical application may be vegetable, synthetic or semisynthetic oils, such as liquid fatty acid esters having in each case 8 to 22 carbon atoms in the fatty acid chain, for example palmitic acid, lauric acid, tridecanoic acid, heptadecanoic acid, stearic acid, arachidic acid, myristic acid, behenic acid, pentadecanoic acid, linoleic acid, elaidic acid, brassic acid, erucic acid or oleic acid, esterified with mono- to trihydric alcohols having 1 to 6 carbon atoms, such as methanol, ethanol, propanol, butanol, pentanol or isomers thereof, ethylene glycol or glycerol. The example of such fatty acid ester is caprylic acid / capric acid ester, polyoxyethylene triolein, ethyl oleate, waxy fatty acid ester such as artificial duck tail gland fat, coconut fatty acid isopropyl ester, oleyl oleate, decyl oleate, ethyl lactate, dibutyl phthalate, diisopropyl adipate, polyol fatty acid ester etc. of commercially available miglyols, isopropyl myristate, isopropyl palmitate, isopropyl stearate, PEG 6-capric acid, saturated fatty alcohol.It is also suitable that the silicone oil of different viscosities or fatty alcohol such as isotridecanol, 2-octyldodecanol, cetyl stearyl alcohol or oleyl alcohol or fatty acid such as oleic acid.In addition, it is also possible to use vegetable oil, such as castor oil, almond oil, olive oil, sesame oil, cottonseed oil, peanut oil, soybean oil etc.
[0046] Suitable solvents, gelling agents and solubilizers are water or water-miscible solvents. Examples of suitable substances are alcohols such as ethanol or isopropanol, benzyl alcohol, 2-octyldodecanol, polyethylene glycols, phthalates, adipic acid esters, propylene glycol, glycerol, dipropylene glycol or tripropylene glycol, waxes, methyl cellosolve, cellosolve, esters, morpholines, dioxane, dimethyl sulfoxide, dimethylformamide, tetrahydrofuran, cyclohexanone, etc.
[0047] The mixture of gelling agent and film former is also entirely possible. In this case, ionic macromolecules such as sodium carboxymethylcellulose, polyacrylic acid, polymethacrylic acid and its salt, sodium amylopectin hemiglycolate, alginic acid or propylene glycol alginate such as sodium salt, gum arabic, xanthan gum, guar gum or carrageenan are particularly used. The following can be used as additional formulation adjuvants: glycerol, paraffin of different viscosities, triethanolamine, collagen, allantoin and novantisolic acid. The formulation may also require the use of surfactants, emulsifiers or wetting agents, such as sodium lauryl sulfate, fatty alcohol ether sulfate, N-lauryl-iminodipropionic acid disodium, polyethoxylated castor oil or sorbitan monooleate, sorbitan monostearate, polysorbate (e.g., Tween), cetyl alcohol, lecithin, glyceryl monostearate, polyoxyethylene stearate, alkylphenol polyethylene glycol ether, cetyl trimethylammonium chloride or mono- / dialkyl polyethylene glycol ether orthophosphoric acid monoethanolamine salt. Stabilizers for stabilizing emulsions, such as montmorillonite or colloidal silicic acid, or stabilizers for preventing decomposition of active substances, such as antioxidants, for example tocopherol or butylhydroxyanisole, or preservatives, such as parabens, can likewise be used to prepare the desired formulations.
[0048] Preparations for parenteral administration may be in the form of individual dosage units, such as ampoules or vials. Solutions of the active compound are preferably used, preferably aqueous solutions, in particular isotonic solutions and suspensions. These injection forms can be prepared as ready-to-use preparations or can be prepared directly before use by mixing the active compound (e.g. lyophilized powder, where appropriate, containing other solid carrier substances) with the desired solvent or suspending agent.
[0049] Intranasal preparations may be present as aqueous or oily solutions or as aqueous or oily suspensions. They may also be present as lyophilized powders prepared using appropriate solvents or suspending agents before use.
[0050] The inhalable preparation may be present as a powder, solution or suspension. Preferably, the inhalable preparation is in powder form, for example as a mixture of the active ingredient with a suitable formulation aid such as lactose.
[0051] The products are manufactured, packaged and sealed under conventional antimicrobial and aseptic conditions.
[0052] As noted above, the compounds of the invention can be administered as a combination therapy with other active agents, such as therapeutically active compounds useful in the treatment of cancer (e.g., prostate cancer, ovarian cancer, lung cancer, or breast cancer). For combination therapy, the active ingredients can be formulated as a composition containing several active ingredients in a single dosage form and / or as a kit containing individual active ingredients in separate dosage forms. The active ingredients used in combination therapy can be administered together or separately.
[0053] It should be understood that the total daily dosage of the compounds and compositions described herein can be determined by the attending physician within the scope of sound medical judgment. The specific therapeutically effective dosage level for any particular patient will depend on a variety of factors, including the disorder being treated and the severity of the disorder; the activity of the specific compound used; the specific composition used; the patient's age, weight, general health, sex and diet; the time of administration and excretion rate of the specific compound used, the duration of treatment, drugs used in combination or concomitantly with the specific compound used; and similar factors well known to researchers, veterinarians, doctors or other clinicians with ordinary skills.
[0054] Depending on the route of administration, a wide range of permissible doses are contemplated herein, including doses falling within the range of from about 1 μg / kg to about 1 g / kg. The doses may be single or divided, and may be administered according to a variety of dosing schedules, including once a day (qd), twice a day (bid), three times a day (tid), or even once every other day, once a week, once a month, etc. In each case, the therapeutically effective amount described herein corresponds to the example of administration, or alternatively corresponds to the total daily, weekly or monthly dose.
[0055] As used herein, the term "therapeutically effective amount" refers to the amount of an active compound or pharmaceutical agent that elicits a biological or medical response in a tissue system, animal or human that is being sought by a researcher, veterinarian, physician or other clinician, including alleviation of symptoms of the disease or disorder being treated. In one aspect, the therapeutically effective amount is an amount that can treat or alleviate a disease or disease symptom at a reasonable benefit / risk ratio applicable to any medical treatment.
[0056] As used herein, the term "therapeutically effective amount" refers to the amount to be administered to a patient and can be based on body surface area, patient weight, and / or patient condition. In addition, it should be understood that there is a correlation between doses determined for humans and those determined for animals (including experimental animals) (exemplarily based on milligrams per square meter of body surface), as described in Freireich, EJ, et al., Cancer Chemother. Rep. 1966, 50 (4), 219 (the disclosure of which is incorporated herein by reference). Body surface area can be approximately determined from patient height and weight (see, e.g., Scientific Tables, Geigy Pharmaceuticals, Ardley, New York, pp. 537-538 (1970)). A therapeutically effective amount of a compound described herein can be defined as any amount that can be used to inhibit the growth of (or kill) a population of malignant cells or cancer cells, such as can be found in a patient in need of relief from such a cancer or malignancy. Typically, such an effective amount ranges from about 5 mg / kg to about 500 mg / kg, from about 5 mg / kg to about 250 mg / kg, and / or from about 5 mg / kg to about 150 mg / kg of the compound / patient body weight. It should be understood that the effective dose may also vary with the route of administration, the use of optional excipients, and the possibility of co-use of the compound with other conventional and non-conventional therapeutic treatments (including other anti-tumor agents, radiation therapy, etc.).
[0057] The term "patient" as used herein includes humans and non-human animals such as companion animals (dogs, cats, etc.) and livestock animals. Livestock animals are animals raised for food production. The patient to be treated is preferably a mammal, particularly a human.
[0058] PD-1 programmed cell death-1; PD-L1 programmed death-ligand-1; GNN graph neural network; EGNN energy graph neural network; SVM support vector machine; RF random forest; FDA, U.S. Food and Drug Administration; HTRF, homogeneous time-resolved fluorescence; PPh3, triphenylphosphine; DIAD, diisopropyl azodicarboxylate; THF, tetrahydrofuran; Me4Si tetramethylsilane; ML, machine learning; AUROC, area under the receiver operating characteristic curve.
[0059] The present invention generally relates to novel compounds for therapeutic use. In particular, the present disclosure relates to novel compounds with immunomodulatory activity that can be used to treat various cancers.
[0060] Also described herein are pharmaceutical compositions of such compounds and methods of treating cancer patients by administering therapeutically effective amounts of such compounds alone, together with other therapeutic agents, or in pharmaceutical compositions.
[0061] In certain exemplary embodiments, the invention relates to a pharmaceutical composition comprising one or more compounds as disclosed herein, or pharmaceutically acceptable salts thereof, and one or more diluents, excipients, or carriers.
[0062] In certain exemplary embodiments, the present invention relates to a compound having formula (I) or a pharmaceutically acceptable salt thereof:
[0063]
[0064] in
[0065] Ar1 is optionally substituted aryl or heteroaryl;
[0066] R and R are independently hydrogen, halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, thiolalkyl, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted;
[0067] R is halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted; and
[0068] Q is:
[0069]
[0070] or one or several amino acid residues.
[0071] In certain exemplary embodiments, the present invention relates to compounds of formula (I) as disclosed herein, wherein Ar1 is phenyl, 2,3-dihydrobenzo[ b [1,4]-dioxine or phenyl(thiazol-2-yl)methanol.
[0072] In certain exemplary embodiments, the present invention relates to compounds of formula (I) as disclosed herein, wherein R1 and R2 are independently hydrogen, methyl, hydroxy, methoxy or -OCH2Ar.
[0073] In certain exemplary embodiments, the invention relates to compounds of formula (I) as disclosed herein, wherein R3 is CH3, CN or Cl.
[0074] In certain exemplary embodiments, the present invention relates to a compound of formula (I) as disclosed herein, wherein the compound comprises
[0075]
[0076] In certain other exemplary embodiments, the present invention relates to a compound having formula (II) or a pharmaceutically acceptable salt thereof:
[0077]
[0078] in
[0079] R1 is aryl, substituted aryl, heteroaryl;
[0080] R is a primary or secondary amine containing alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted;
[0081] R is halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted; and
[0082] Ar1 is aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted.
[0083] In certain exemplary embodiments, the present invention relates to compounds of formula (II) as disclosed herein, wherein Ar1 is phenyl, 2,3-dihydrobenzo[ b [1,4]-dioxine or phenyl(thiazol-2-yl)methanol.
[0084] In certain exemplary embodiments, the present invention relates to compounds of formula (II) as disclosed herein, wherein R1 and R2 are independently piperidine, pyrrolidine, phenyl, 4-halophenyl (halo = fluoro, bromo, iodo) and / or one or more amino acid residues as a single amino acid or a combination of amino acids.
[0085] In certain exemplary embodiments, the present invention relates to compounds of formula (II) as disclosed herein, wherein R3 is methyl, CN or halogen.
[0086] In certain exemplary embodiments, the present invention relates to a compound of formula (II) as disclosed herein, wherein the compound is
[0087] .
[0088] In certain exemplary embodiments, the present invention relates to a compound having formula (III) or a pharmaceutically acceptable salt thereof:
[0089]
[0090] in
[0091] Ar1 is phenyl, 2,3-dihydrobenzo[ b [1,4]dioxine or phenyl(thiazol-2-yl)methanol;
[0092] Ar2 is piperidine or pyrrolidine;
[0093] Ar3 is phenyl, 4-halophenyl (halo = fluoro, bromo, iodo); and
[0094] X is independently halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted.
[0095] In certain exemplary embodiments, the invention relates to compounds of formula (III) as disclosed herein, wherein X is methyl, cyano or chloro.
[0096] In certain exemplary embodiments, the invention relates to a compound of formula (III) as disclosed herein, wherein the compound is
[0097] .
[0098] In certain exemplary embodiments, the present invention relates to a pharmaceutical composition comprising one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more diluents, excipients or carriers.
[0099] In certain exemplary embodiments, the invention relates to one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more diluents, excipients or carriers, wherein the compound is an immunomodulator.
[0100] In certain exemplary embodiments, the present invention relates to one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more diluents, excipients or carriers, wherein the compound is an inhibitor of PD-1 and PDL-1 signaling pathways.
[0101] In certain exemplary embodiments, the invention relates to one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more diluents, excipients or carriers, wherein the compounds are used to treat cancer.
[0102] In certain exemplary embodiments, the present invention relates to a method for treating a patient having cancer, comprising the steps of administering to a patient in need of relief from said cancer a therapeutically effective amount of one or more compounds of formula (I), (II), (III) or a pharmaceutically acceptable salt thereof, and one or more carriers, diluents or excipients.
[0103] In certain exemplary embodiments, the present invention relates to a method for treating a patient having cancer, comprising the steps of administering to a patient in need of relief from said cancer a therapeutically effective amount of one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more carriers, diluents or excipients, wherein said compound is an immunomodulator.
[0104] In certain exemplary embodiments, the present invention relates to a method for treating a patient having cancer, comprising the steps of administering to a patient in need of remission from said cancer a therapeutically effective amount of one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more other compounds having the same or different mode of action, and one or more carriers, diluents or excipients, wherein said compound is an immunomodulator.
[0105] In certain exemplary embodiments, the present invention relates to a method for treating a patient having cancer, comprising the steps of administering a therapeutically effective amount of one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more carriers, diluents or excipients to a patient in need of remission from said cancer, wherein said cancer is castration-resistant prostate cancer.
[0106] In certain exemplary embodiments, the present invention relates to a method for treating a patient having cancer, comprising the steps of administering to a patient in need of relief from said cancer a therapeutically effective amount of one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more carriers, diluents or excipients, wherein said compound is an immunomodulator.
[0107] In certain exemplary embodiments, the present invention relates to a pharmaceutical composition for use as a cancer medicament, comprising one or more compounds of formula (I), (II), (III) or a pharmaceutically acceptable salt thereof, and one or more diluents, excipients or carriers.
[0108] In certain exemplary embodiments, the present invention relates to a drug conjugate comprising one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more carriers, diluents or excipients.
[0109] In certain exemplary embodiments, the invention relates to a drug conjugate comprising one or more compounds of formula (I), (II), (III) or pharmaceutically acceptable salts thereof, and one or more carriers, diluents or excipients, wherein the conjugate confers cell type or tissue type targeting or the conjugate targets another pathway that synergizes with the effects of those compounds.
[0110] In certain exemplary embodiments, the present invention relates to a method for treating a cancer patient, comprising the steps of administering to a patient in need of remission from said cancer a therapeutically effective amount of one or more compounds, and one or more carriers, diluents or excipients, wherein said compound has formula (I), (II) or (III).
[0111] In certain exemplary embodiments, the invention relates to a compound having formula (IV) or a pharmaceutically acceptable salt thereof:
[0112]
[0113] in
[0114] represents a single bond or a double bond;
[0115] represents an optional cyclic ring;
[0116] R is hydrogen, halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted;
[0117] R is hydrogen, halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted;
[0118] and
[0119] X is carbon or nitrogen.
[0120] In certain exemplary embodiments, the invention relates to a compound of formula (IV) as disclosed herein, wherein the compound is
[0121]
[0122]
[0123] .
[0124] In certain exemplary embodiments, the invention relates to a compound having formula (V) or a pharmaceutically acceptable salt thereof:
[0125]
[0126] in
[0127] R1, R2 and R3 independently represent five substituents selected from hydrogen, halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted; or any two adjacent substituents are linked together to form a cyclic or heterocyclic moiety.
[0128] In certain exemplary embodiments, the invention relates to compounds of formula (V) as disclosed herein, wherein R1 = CH3.
[0129] In certain exemplary embodiments, the invention relates to compounds of formula (V) as disclosed herein, wherein R1 = Cl.
[0130] In certain exemplary embodiments, the invention relates to a compound of formula (V) as disclosed herein, wherein the compound is
[0131]
[0132]
[0133] .
[0134] In certain exemplary embodiments, the invention relates to a compound having Formula VI or VII, or a pharmaceutically acceptable salt thereof:
[0135]
[0136] in
[0137] A is carbon or nitrogen;
[0138] L is (CH2) n , -SO, -SO2, -CO, -CO(CH2)O, where n is 0, 1, 2;
[0139] Ar1 is aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted;
[0140] R is halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted; and
[0141] R2 is H, methyl, ethyl or any alkyl group;
[0142] R3 is halogen, -OR, -NO2, cyano, -NR a R b 、-N3、-S(O)2R a , -C(alkyl), -C(cycloalkyl), C(alkynyl), C(haloalkyl), acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted, wherein R, R a and R b are independently alkyl;
[0143] R4 is halogen, amino acid, saturated or unsaturated aromatic or heteroaromatic ring, carbohydrate derivative; or -(CH2) m NR a R b , where m = 0, 1, 2; where R a and R b are independently alkyl;
[0144] or R3 and R4 are linked together to form a cyclic or heterocyclic moiety; and
[0145] R5 is halogen, -OR, -NO2, cyano, -NR a R b 、-N3、-S(O)2R a , -C(alkyl), -C(cycloalkyl), C(alkynyl), C(haloalkyl), acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted, wherein R, R a and R b are independently alkyl.
[0146] In certain exemplary embodiments, the invention relates to a compound having Formula VIII or a pharmaceutically acceptable salt thereof:
[0147]
[0148] in
[0149] n is 0, 1, or 2;
[0150] A is carbon or nitrogen;
[0151] R is halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted; and
[0152] R2 is independently halogen, -OR, -NO2, cyano, -NR a R b 、-N3、-S(O)2R a , -C(alkyl), -C(cycloalkyl), C(alkynyl), C(haloalkyl), acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted, wherein R, R a and R b are independently alkyl, wherein R, R a and R b are independently alkyl;
[0153] or R2 and R3 are linked together to form a cyclic or heterocyclic moiety;
[0154] R3 is independently -(CH2) m NR a R b : wherein m=0 to 2, halogen, any amino acid, any saturated or unsaturated aromatic or heteroaromatic ring or carbohydrate derivative, wherein R a and R b are independently alkyl;
[0155] R4 is independently halogen, -OR, -NO2, cyano, -NR a R b 、-N3、-S(O)2R a , -C(alkyl), -C(cycloalkyl), C(alkynyl), C(haloalkyl), acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted, wherein R, R a and R b are independently alkyl, wherein R, R a and R b are independently alkyl; and
[0156] Ar1 is aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted.
[0157] In certain exemplary embodiments, the invention relates to a compound having Formula IX or a pharmaceutically acceptable salt thereof:
[0158]
[0159] in
[0160] n is 0, 1, or 2;
[0161] R is halogen, azido, nitro, cyano, alkyl, alkenyl, alkynyl, alkylalkynyl, alkoxy, hydroxyalkyl, aminoalkyl, alkylthio, mercaptoalkyl, heteroalkyl, heteroalkenyl, heteroalkynyl, heterocyclyl, cycloalkyl, cycloalkenyl, cycloheteroalkyl, cycloheteroalkenyl, acyl, aryl, heteroaryl, arylalkyl, arylalkenyl, or arylalkynyl, each of which is optionally substituted; and
[0162] R2 is independently halogen, -OR, -NO2, cyano, -NR a R b 、-N3、-S(O)2R a, -C(alkyl), -C(cycloalkyl), C(alkynyl), C(haloalkyl), acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted, wherein R, R a and R b are independently alkyl;
[0163] R3 is independently -(CH2) m NR a R b : wherein m=0 to 2, halogen, any amino acid, any saturated or unsaturated aromatic or heteroaromatic ring or carbohydrate derivative, wherein R a and R b are independently alkyl;
[0164] R4 is independently halogen, -OR, -NO2, cyano, -NR a R b 、-N3、-S(O)2R a , -C(alkyl), -C(cycloalkyl), C(alkynyl), C(haloalkyl), acyl, aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted, wherein R, R a and R b are independently alkyl;
[0165] and
[0166] Ar1 is aryl, heteroaryl, arylalkyl, arylalkenyl or arylalkynyl, each of which is optionally substituted.
[0167] In certain exemplary embodiments, the present invention relates to a pharmaceutical composition comprising one or more compounds of formula (IV) to (IX) as disclosed herein or pharmaceutically acceptable salts thereof and one or more diluents, excipients or carriers.
[0168] In certain exemplary embodiments, the present invention relates to a pharmaceutical composition comprising one or more compounds of formula (IV) to (IX) as disclosed herein or pharmaceutically acceptable salts thereof and one or more diluents, excipients or carriers, wherein the compound is an immunomodulator.
[0169] In certain exemplary embodiments, the present invention relates to a pharmaceutical composition comprising one or more compounds of formula (IV) to (IX) as disclosed herein or pharmaceutically acceptable salts thereof and one or more diluents, excipients or carriers, wherein the compound is an inhibitor of the PD-1 and PDL-1 signaling pathways.
[0170] In certain exemplary embodiments, the present invention relates to a method for treating a cancer patient, the method comprising the steps of administering a therapeutically effective amount of one or more compounds of formula (IV) to (IX) as disclosed herein, or a pharmaceutically acceptable salt thereof, and one or more diluents, excipients, or carriers.
[0171] In certain exemplary embodiments, the present invention relates to a pharmaceutical composition comprising one or more compounds of formula (IV) to (IX) as disclosed herein or pharmaceutically acceptable salts thereof and one or more diluents, excipients or carriers, wherein the compound is an inhibitor of the PD-1 and PDL-1 signaling pathways.
[0172] In certain exemplary embodiments, the present invention relates to a method for treating a cancer patient, the method comprising the steps of administering to a cancer patient in need of remission from the cancer a therapeutically effective amount of one or more compounds of formula (IV) to (IX) as disclosed herein and one or more carriers, diluents or excipients.
[0173] In certain exemplary embodiments, the present invention relates to a method for treating a cancer patient, the method comprising the steps of administering to a cancer patient in need of remission from the cancer a therapeutically effective amount of one or more compounds of formula (IV) to (IX) as disclosed herein and one or more other compounds having the same or different mode of action and one or more carriers, diluents or excipients.
[0174] In certain exemplary embodiments, the present invention relates to a method for treating a cancer patient, the method comprising the steps of administering a therapeutically effective amount of one or more compounds of formula (IV) to (IX) as disclosed herein and one or more carriers, diluents or excipients to a cancer patient in need of remission from the cancer, wherein the compound is an inhibitor of the PD-1 and PDL-1 signaling pathways.
[0175] In certain exemplary embodiments, the present invention relates to a method for treating a cancer patient, the method comprising the steps of administering a therapeutically effective amount of one or more compounds of formula (IV) to (IX) as disclosed herein and one or more carriers, diluents or excipients to a cancer patient in need of remission from the cancer, wherein the cancer is castration-resistant prostate cancer.
[0176] In certain exemplary embodiments, the present invention relates to pharmaceutical compositions for use as cancer medicaments, comprising one or more compounds of Formula (I) to (IX) as disclosed herein and one or more carriers, diluents or excipients.
[0177] The programmed cell death protein 1 / programmed death-ligand 1 (PD-1 / PD-L1) interaction is an immune checkpoint used by cancer cells to enhance immunosuppression. There is an urgent need to develop small molecule drugs that are fast-acting, cost-effective, and easily bioavailable compared to antibodies. Unfortunately, it is time-consuming and expensive to synthesize and validate large libraries of small molecules that inhibit the PD-1 / PD-L1 interaction in a blind manner. To improve this drug discovery pipeline, we have developed a machine learning approach trained on patent data to identify, synthesize, and validate PD-1 / PD-L1 small molecule inhibitors. Our model contains two features: a docking score representing binding energy (E) as a global feature, and a subgraph feature through a graph neural network (GNN) of molecular topology as a local feature. This energy-graph neural network (EGNN) model outperformed traditional machine learning methods as well as a simple GNN (with an F1 score of 0.9524 and a Cohen's kappa score of 0.8861 for the holdout test set), suggesting that the topology of the small molecule, the structural interactions in the binding pocket, and the chemical diversity of the training data are all important considerations for enhancing model performance. The bootstrapped EGNN model was used to select compounds for synthesis and experimental validation of predicted high and low potency for inhibiting PD-1 / PD-L1 interaction. The potent inhibitor (4-((3-(2,3-dihydrobenzo[b][1,4]dioxin-6-yl)-2-methylbenzyl)oxy)-2,6-dimethoxybenzyl)-D-serine is a hybrid of two known bioactive scaffolds and has an IC50 value of 339.9 nM, which is relatively superior to known bioactive compounds. We conclude that our bootstrapped EGNN model will be useful for identifying target-specific and highly potent molecules designed by scaffold hopping, a well-known medicinal chemistry technique.
[0178] Traditionally, the development of small molecule inhibitors requires high-throughput screening of large libraries of multi-drug class compounds. 10 or medicinal chemists’ iteration of scaffolds with weak receptor activity to enhance potency 11 This entire process is (i) time-consuming; (ii) requires expensive instruments and robotics; (iii) based on trial-and-error methods; and (iv) very inefficient for rapidly identifying several new scaffolds. 12 In addition, virtual screening using docking methods has been developed to improve the process, but with limited success. 13 In addition, ML architectures such as support vector machines (SVM) 14-16 , Random Forest (RF) 17-19 , Graph Convolutional Network 20 and Graph Neural Networks (GNNs) 21,22 Already used in drug design and prediction of drug-target interactions 23,24Recently, new architectures that exploit combinations of graph features in protein binding sites have shown great promise in calculating binding affinity and determining whether a compound will bind to a target. 20,22 .
[0179] Several new neural network-based architectures have also been proposed that show promise for identifying effective skeletons, but many have not yet been experimentally tested. 15,16,25-28 , and advances in the ability to mine and characterize protein crystallographic data will hopefully advance the creation of these models. 29 Recently, it has been shown that molecular subgraph features incorporated by GNNs and protein features encoded by their sequences can be combined to predict whether a compound can target a given protein 24 Inspired by this work and based on our interest in developing drug design and immunology approaches 29-36 In this paper, we have developed a novel machine learning model to predict whether a compound can inhibit the PD-1 / PD-L1 interaction. Our approach replaces protein sequence features with a docking score that represents the free energy of binding, and because of this global energy interaction of small molecules in the binding pocket, we have called this model an "Energy Graph Neural Network" (EGNN). Using CANDOCK 31 Calculate the three-dimensional atomic interaction energy score ( Figure 1 B), and use an end-to-end training method with local molecular graph features ( Figure 1 A) Combination Figure 1 CD). In this work, we used this EGNN model to select synthetic designs and experimentally tested a curated list of compounds from these predictions to prospectively identify potent PD-L1 small molecule inhibitors using a homogeneous time-resolved fluorescence (HTRF) assay. We also tested negative predictions, showing that the model can be used to select potent leads as PD-1 / PD-L1 inhibitors.
[0180] Patent data used to train the EGNN model
[0181] We trained our model using PD-1 / PD-L1 small molecule inhibition data of 762 compounds from four patents: WO 2015 / 034820 A1 from BMS 7 and WO 2015 / 160641 A2 (674 compounds) 6 , and WO 2018 / 119263 A1 of Incyte Corporation 37 and US 2018 / 0273519 A1 38(88 compounds). In the patent, the activity against PD-1 / PD-L1 interaction was shown using a homogeneous time-resolved fluorescence (HTRF) binding assay. However, the patent does not list the individual IC values for all compounds. 50 values, but rather provide a range of inhibition for different molecules. Therefore, we trained binary classifiers with cutoff values for both datasets to classify a molecule as either “highly effective” or “lowly effective” ( Figure 1 ). If the IC50 of a molecule reported in the patent is less than or equal to 100 nM, it is considered a “highly effective” molecule, otherwise it is considered a “lowly effective” molecule. This threshold was chosen because it is the only common threshold in the four patents (Table S4). It should be noted that the actual value of IC50 should not be considered here because our multiple repeated experiments could not obtain accurate reported results for some molecules in the patent (see the IC50 value of compound 4a in Table 2, BMS-1 annotated with 6-100 nM in WO 2015 / 034820 A1 patent 7). Therefore, we consider the positive prediction (highly effective) based on our experimental IC50 value compared with the upper limit of the BMS control molecule (compound 4a / BMS-1) in WO 2015 / 034820 A1 patent 7. The training data set of 762 small molecules with BMS or Incyte annotations is shown in the Supporting Information File (TrainingData.xlsx).
[0182] We selected BMS and Incyte patents to include the chemical diversity of the molecules in the training dataset. Figure 2A The distribution of low-potency and high-potency molecules and general frameworks in the BMS and Incyte patents is shown. The BMS patent has 372 high-potency compounds and 302 low-potency compounds, respectively, while the Incyte patent has 47 high-potency compounds and 41 low-potency compounds, respectively. The BMS patent framework contains 417 derivatives of (2-methyl-3-biphenyl)methanol and 257 derivatives of [3-(2,3-dihydro-1,4-benzodioxin-6-yl)-2-methylphenyl]methanol, such as Figure 2A (lower left), the R groups are CN, Cl, Br, and CH3. On the other hand, the Incyte patented framework has different sub-frameworks, Figure 2AIn the figure, the X represents an N or CR group (R: alkyl group) in the Incyte framework. These frameworks suggest that the chemical diversity of Incyte compounds is higher than that of BMS compounds because the general structure of compounds in Incyte patents has more structural diversity in terms of sub-frameworks and atoms. We verified this observation using pairwise Tanimoto similarity scores of BMS and Incyte compounds, respectively. Figure 2B and 2C The pairwise Tanimoto similarity was calculated using the Morgan fingerprint with a radius of 2 and a bit length of 1024. The high red area in the BMS heat map indicates that the molecular pairs are structurally similar to each other. The low red area in the Incyte heat map indicates that it has more chemical diversity in terms of molecular structure. Moreover, the average pairwise Tanimoto40 similarity scores for all BMS compounds were found to be 0.4434 and 0.3920 for all Incyte compounds, respectively, which confirms the higher chemical diversity in Incyte compounds compared to BMS compounds.
[0183] PD-L1 homodimer and PD-1 / PD-L1 crystal structures reveal docking binding sites
[0184] It has been previously shown that BMS compounds inhibit PD-1 / PD-L1 interaction by inducing dimerization of PD-L1 8,9 Therefore, the PD-L1 homodimer crystal structure (PDB ID: 5N2F) was selected for docking all compounds in this manuscript. The PyMol software package was also used 40 The PD-1 / PD-L1 crystal structure (PDB ID: 4ZQK) was used to check whether the binding sites of PD-L1 in the homodimer crystal structure (5N2F) overlapped and aligned with each other ( Figure 3A ).exist Figure 3B In the figure, the selected binding site of the PD-L1 homodimer on the superimposed and aligned crystal structures is shown to indicate that the formation of homodimers of PD-L1 with small molecules blocks the PD-1 / PD-L1 interaction. Known inhibitors of the PD-1 / PD-L1 interaction in the selected binding site (ligand ID: 8HW) 8 ( Figure 3C) suggested that the selected binding site corresponding to the PD-L1 homodimer is relevant for the development of PD-1 / PD-L1 inhibitors. Therefore, docking interactions of the PD-L1 homodimer will be relevant for the identification of PD-1 / PD-L1 inhibitors. In addition, direct docking with PD-1 / PD-L1 was not performed because the binding site between PD-1 and PD-L1 is filled with interacting amino acid residues from both proteins. Therefore, there is no space to dock small compounds with the PD-1 / PD-L1 complex.
[0185] Using Candock 31 Docking conformations of small molecules to PD-L1 homodimers were generated (for details, see the Experimental section on using docking and energy vectors (E) to generate energy features in EGNN). Before developing the machine learning approach, we also evaluated 96 potential energy scoring functions in CANDOCK using only the docking scores of the compounds in the training set. 31 The ability of each of the two categories to classify efficient (vs) inefficient molecules was evaluated. The Cohen's κ score was used to select the best scoring function to distinguish between the two categories (Table S1). The scoring function radial cumulative complete 15 (RCC15) obtained the highest Cohen's κ score of 0.41447. However, the RCC15 score cannot clearly distinguish all the efficient and inefficient classified molecules in the training data (see violin plots in Figure S1). Using only one scoring function is not enough to capture the different states of PD-1 / PD-L1 inhibition of small molecules. Therefore, we developed an EGNN model using the top scoring function of each category, which exhibited a positive Cohen's κ value (Table S1) to represent the global feature (see the Experimental Section on Generating Energy Features Using Docking and Energy Vectors (E) in EGNN for details). This includes the RCR15 (κ= 0.37746) and RCC15 (κ=0.41447) scoring functions. Models with a κ score between 0.21 and 0.40 were considered to have fairly good consistency, and if the κ score was between 0.41 and 0.60, the model was considered to have moderate consistency. 42 .
[0186] The EGNN model with hyperparameter optimization outperforms GNN and other baseline models
[0187] A detailed description of the EGNN model including the combination of a molecular GNN with docking is given in the experimental section. Figure 1The EGNN model is shown to be a combination of local features of small molecules represented as GNNs (see Graph Neural Networks for Molecular Graphs in EGNN) and global features of protein-ligand interactions represented as docking scores (see Section Generating Energy Features Using Docking and Energy Vectors (E) in EGNN). EGNN was trained using 88 small molecules with high and low efficacy for inhibition of PD-1 / PD-L1 from two Incyte patents (see Patent Data for Training EGNN Model). We calculated the change in the average F1 score (over five cross-validation folds) with respect to the number of epochs for different hyperparameters (Figure S2). The optimal hyperparameters were selected to avoid overfitting and underfitting of EGNN, including: dimension (dim) of hidden molecule vector = 10, subgraph radius = 2, and number of hidden layers = 1 (see Experimental section on EGNN training and hyperparameter optimization).
[0188] The EGNN and GNN models were trained using different training sets to examine the impact of chemical diversity on the model performance in the classification of highly effective and ineffective molecules. Two datasets (BMS and Incyte) were used separately and in combination to train the EGNN model and determine the best dataset for predicting PD-1 / PD-L1 inhibitors. The datasets were split into training-validation and test sets (4:1) using two different methods: (1) using a random splitter on shuffled data; and (2) using a skeleton splitting method using the DeepChem library. 43 . The models were then trained with five-fold cross validation and the test set was used to evaluate the predictive ability of the models. Here, Cohen's κ, F1 score, and area under the receiver operating characteristic curve (AUROC) were measured to compare the three models trained with BMS data only, Incyte data only, and BMS-Incyte combined data. In addition, as a separate experiment, all measurements were obtained for EGNN and GNN models trained only on BMS data while predicting for Incyte data and vice versa.
[0189] Figure 4A shows how to split the dataset in the training-validation-test set scheme. The initial dataset is divided into two groups with a 4:1 ratio based on skeleton splitting or random shuffle splitting. Then 80% of the dataset is used as training and validation datasets, while 20% of the dataset is used as a reserved test set to evaluate model performance.
[0190] Figure 4A shows the average F1 scores (after five-fold cross validation) for two models trained using BMS compounds, Incyte compounds, and the union of these sets. The average F1 scores for the EGNN and GNN models trained using the Incyte data were 0.956 (±0.051) and 0.678 (±0.157), respectively (Figure 4A). This result suggests that the use of a variety of chemical skeletons ( Figure 2C ) performed much better than the GNN trained with the same dataset. However, when the same experiment was performed only with BMS compounds with less chemical diversity than Incyte, the average F1 score was comparable for both models, 0.992 (±0.007) for the EGNN model and 0.948 (±0.022) for the GNN model. This suggests that the GNN model performs well with less chemical diversity in the training data than with greater chemical diversity in the training data. However, the EGNN model performed well in both datasets, indicating that it is a model superior to the GNN.
[0191] The GNN model with backbone splitting appears to produce comparable results to EGNN (Figure S4). However, this is expected since graph neural networks use two-dimensional molecular frameworks / topologies in training. GNNs perform well when the framework distribution of compounds is similar in the training-validation and test sets. However, our intention is to develop a model that can be used to screen large libraries of compounds that do not need to share the same backbone distribution as the training set (i.e., Incyte or BMS). Therefore, we chose a random splitter with shuffling to create a test set for performance evaluation to develop a more general model.
[0192] For both models trained with the BMS compound, the Incyte compound, and the union of these sets, the Cohen's κ scores for the different test sets (the holdout test set was based on the random split) are as follows: Figure 4B The κ scores of the EGNN and GNN models trained with the Incyte data and tested on the holdout test set were 0.8861 and 0.4304, respectively ( Figure 4B ). This result shows that the use of a variety of chemical skeletons ( Figure 2C) performed much better than the GNN trained with the same dataset. However, when the same experiment was performed only with BMS compounds, which have lower chemical diversity than Incyte, the Cohen's κ scores of the two models were comparable, 0.6416 for the EGNN model and 0.7164 for the GNN model. This suggests that the GNN model performs well when the chemical diversity in the training and test data is smaller than that in the larger chemical diversity. The two models also showed comparable performance for the combined dataset. When combining both the BMS and Incyte datasets, the κ score of the EGNN model for the retained test set was 0.6072, while that of the GNN model was 0.6729. Comparing the three different training sets, similar trends were observed for the F1 score ( Figure 4C ). These results suggest that the EGNN model outperforms the GNN model on chemically diverse datasets such as the Incyte data. We believe this is due to the addition of the “global” energy features captured by the docking score of the PD-L1 homodimer as training data in the EGNN, compared to the only “local” structural features of the small molecules in the GNN model training data.
[0193] We also investigated the ability of EGNN and GNN models trained on one compound set to predict highly and poorly effective inhibitors of PD-1 / PD-L1 in another compound set. Figure 4B and 4C (κ score and F1 score, respectively), with different bars representing different test sets. The Tanimoto similarity between Incyte and BMS compounds is also shown Figure 4DHeat map. The average pairwise Tanimoto similarity score of 0.3044 indicates that the compounds within the two datasets are very different from each other. When the EGNN and GNN models were trained on BMS compounds and used to predict Incyte compounds, Cohen's κ scores of 0.1505 for EGNN and 0.1200 for GNN were observed, and F1 scores of 0.3810 and 0.2264, respectively. On the other hand, when they were trained using Incyte data and used to predict BMS compounds, both models showed improvements in F1 and κ scores (κ scores for EGNN = 0.3852, GNN = 0.3196 and F1 scores for EGNN = 0.7400, GNN = 0.6958). These results show that both the F1 score and Cohen's κ score of the EGNN and GNN models improved significantly when trained with Incyte data and tested on BMS. However, the AUROC score was unable to correctly distinguish between these models (Figure 3). This is expected since AUROC is not a good metric for evaluating models trained on skewed / imbalanced datasets and it may hinder poor performance of the model. 44 . In addition, these results suggest that the EGNN model outperforms the GNN model in both cases (see Table S2 for details). This highlights the importance of chemical diversity in the training data, even if there is not much compound similarity between the training and test sets. Therefore, it is not appropriate to use BMS alone or combined BMS and Incyte data to train the final model for prediction of unknown molecules. Therefore, we only selected the Incyte dataset to train the EGNN model, which significantly improved the EGNN model performance (Table S2).
[0194] Furthermore, we have examined the ability to classify the training set compounds into the low- and high-efficiency classes by comparing Tanimoto 2D similarity alone. Violin plots showing the distribution of Tanimoto 2D similarity scores for low-efficiency, high-efficiency, and all compounds are shown in Figure S5. This clearly shows that compounds in both the high-efficiency and low-efficiency classes show a high probability of having low pairwise similarity scores, and this is true even when all compounds are considered. Therefore, it is not sufficient to consider only topological similarity to select potent PD-1 / PD-L1 inhibitors.
[0195] Finally, we compared the cross-validated EGNN model trained with the Incyte training data with the GNN, support vector machine (SVM), and random forest (RF) baseline models using their test set performance. Both the SVM and RF models were trained on local and global features. The Weisfeiler-Lehman algorithm was used 45Fingerprints extracted from molecular graphs with a radius of 2 were used as “local” features similar to the EGNN and GNN models. Here, we zero-padded to the maximum fingerprint length to keep the same fingerprint size. The same pre-selected docking scores (RCR15 and RCC15) obtained by CANDOCK were used as “global” energy features. For all four models, the AUROC, AUPRC, precision, call, F1 score and Cohen's κ value obtained are listed in Table 1. The scikit-learn library with the “linear” kernel was used. 46 The SVM model was trained using the “svm” package from scikit-learn with 500 trees. 46 The "Random Forest Classifier" in scikit-learn package 46 The "metrics" module in R was used to calculate the AUROC, precision, recall, F1 score, and Cohen's κ statistics. The "precrec" library in R programming language was used 47 The precision recall curves of the models and AUPRC values were obtained. The EGNN model outperformed all other models, with AUROC, AUPRC, precision, call, F1 score and Cohen's κ values of 0.9250, 0.9212, 0.9091, 1.0000, 0.9524 and 0.8861, respectively (Table 1). Comparison of the precision recall curves of the four models (Fig. 4E) also confirmed that the EGNN model was superior to all the other three models. In summary, with the Incyte dataset, the local and global features combined in the EGNN provided the best performance.
[0196] Table 1. AUROC, AUPRC, precision, recall, F1 score, and Cohen's κ of EGNN for PD-1 / PD-L1 inhibitor prediction compared with other baseline models such as random forest, SVM, and GNN models. All models were trained on the Incyte dataset and evaluated on the same held-out test set.
[0197]
[0198] Synthetic selection and validation of EGNN predictions for PD-1 / PD-L1 inhibition
[0199] The EGNN model trained using the optimal hyperparameters and the Incyte dataset was used to obtain predictions for an in-house database of small molecule designs. We developed a bootstrapped EGNN model to predict compounds with high and low efficacy for PD-1 / PD-L1 inhibition using 1000 EGNN models (see the Bootstrapped EGNN Model section). Bootstrapping is an important statistical technique that can be used to select reliable molecules for synthesis and experimental validation based on the consistency between multiple models. The bootstrapped EGNN model identified highly and poorly effective small molecules as PD-1 / PD-L1 inhibitors, which were synthesized and then experimentally validated using the HTRF binding assay (see Table 2 for a summary). Specifically, we selected four molecules predicted to be highly or poorly effective for PD-1 / PD-L1 inhibition for testing based on the bootstrapped EGNN SoftMax mean score and standard deviation (see EGNN SoftMax score in Table 2).
[0200] Among the predictions bootstrapped by EGNN, we have selected 1 highly effective molecule (compound 4b) and 3 inefficient molecules with different scaffolds (compounds 4c, 4d, and 4e) for further testing. We have defined a new parameter called "counts", which records the number of models among the 100 models that give a SoftMax score of 0.5 or higher for the molecule of interest. Specifically, compound 4b is predicted to be a highly effective PD-1 / PD-L1 inhibitor with a count of 99 and an average SoftMax score of 0.7771 (±0.1193). In contrast, the results for compound 4c are only 69 counts and an average SoftMax value of 0.5786 (±0.1406), the results for compound 4d are only 5 counts and an average SoftMax value of 0.1821 (±0.1514), and the results for compound 4e are 62 counts and an average SoftMax value of 0.5280 (±0.1259), indicating inefficient predictions. We also synthesized a BMS backbone (compound 4a, a known PD-1 / PD-L1 inhibitor) to be used as a positive control for our HTRF experiments. The compound structures are shown in Schemes 1 and 2 (see Experimental Section for procedures and characterization). The predicted high potency molecule (compound 4b) is (4-((3-(2,3-dihydrobenzo[ b [1,4]dioxin-6-yl)-2-methylbenzyl)oxy)-2,6-dimethoxybenzyl)-D-serine, a hybrid of two BMS molecules, namely 4a (BMS-1) and BMS-1002 containing (2-methyl-3-biphenyl)methanol and [3-(2,3-dihydro-1,4-benzodioxin-6-yl)-2-methylphenyl]methanol, respectively ( Figure 5A ), and demonstrated the ability of the EGNN model to perform skeleton hopping.
[0201] The EGNN trained with the Incyte data had only three compounds containing the [3-(2,3-dihydro-1,4-benzodioxin-6-yl)-2-methylphenyl]methanol skeleton. As a separate experiment, we have removed these and made predictions for our synthetic library, and the EGNN model was still able to predict compound 4b (a compound based on the [3-(2,3-dihydro-1,4-benzodioxin-6-yl)-2-methylphenyl]methanol skeleton) as a highly potent compound with a SoftMax score of 0.8285 ± 0.1396 and a count of 971. This result suggests that the EGNN model can identify highly potent PD-1 / PD-L1 inhibitors with a [3-(2,3-dihydro-1,4-benzodioxin-6-yl)-2-methylphenyl]methanol skeleton, even without learning on a similar skeleton.
[0202] Scheme 1. Representative synthetic scheme a
[0203]
[0204]
[0205] a Reaction conditions: (i) BH3•THF complex (1.0 M in THF), anhydrous THF, 0℃ to room temperature, 2 days; (ii) PPh3, DIAD, 0℃ to room temperature, 20 h, anhydrous THF; (iii) amine component, NaBH3CN, catalytic amount of AcOH, DMF, 80℃ or room temperature, 1 h or 3 h or overnight.
[0206] Scheme 2. Synthesis of 4d (GCL2)
[0207]
[0208] The top docking pose of the PD-L1 homodimer of compound 4b (PDB ID: 5N2F) interacts in a similar manner as previously shown for the cocrystal structure 8 ( Figure 5B Specifically, for compound 4b, the 2,3-dihydro-1,4-benzodioxine group promotes the amino acid residue Tyr56 ( A It is known that if the 2,3-dihydro-1,4-benzodioxine group is present, the A Tyr56 does not close the hydrophobic pocket from one end 8 This creates a hydrophobic tunnel in the docked conformation ( Figure 5B In addition, the aromatic ring of compound 4b (2,3-dihydro-1,4-benzodioxine) binds to the amino acid residue A The π-π stacking effect of Tyr56 is stabilized ( Figure 5C ). The central methylbenzyl ring in the structure ( Figure 5A The magenta color in 4b) rotates the 2,3-dihydro-1,4-benzodioxine ring by about 30°. o , and the methyl group of the methylbenzyl ring points toward chain B of the PD-L1 homodimer. This orientation results in Met115 of both chains A and B of the homodimer and B Hydrophobic interaction of Ala121. The D-serine terminal of compound 4b with A Thr20 and A Ala121 forms a hydrogen bond and in the 4b molecule A A plausible hydrogen bond is formed between the backbone NH of Tyr123 and the oxygen in one of the two methoxy groups ( Figure 5C ). These results suggest that favorable interactions of compound 4b, which can dimerize PD-L1, will lead to PD-1 / PD-L1 inhibition.
[0209] HTRF assay confirmed that compound 4b inhibited the PD-1 / PD-L1 interaction with an IC 50 is 339.9 nM (see the Experimental section for details) Figure 5D This is higher than the IC of 521.5 nM of BMS compound 4a (BMS-1 molecule in BMS patent WO 2015 / 034820 A1) synthesized and tested in our laboratory. 50 It should be noted that the BMS-1 molecule is indicated in the BMS patent to have an IC of 6-100 nM in the HTRF assay. 50 7 However, multiple repetitions of our experiments did not yield an IC<100 nM for inhibition of PD-1 / PD-L1 interaction. 50 Value (see IC 50 Calculation of IC50 values section and supporting file HTRF_IC50_Data.xlsx). As mentioned before, this result does not affect our machine learning approach, as we have already classified molecules according to high and low efficiency, rather than estimating specific IC 50 One possible explanation for this difference in experimental results between our work and the patent is that the 50 For this reason, we have included a detailed description of the HTRF assay protocol, the protocol used to calculate the IC 50The data are presented in Table 2 and supporting data files for use by the scientific community (see Experimental Section). To test the effectiveness of our bootstrapped EGNN model in correctly identifying poor predictions, we also tested compounds 4c, 4d, and 4e, which resulted in no / poor inhibition of the PD-1 / PD-L1 interaction (Table 2). The IC values of each tested compound were 50 Figure (Figure S6) and 13 C and 1 H NMR spectra are provided as supporting information. The pairwise Tanimoto similarity scores between these 4a-4e (Table S3) show the ability of the EGNN model to identify highly effective and ineffective inhibitors regardless of structural similarity. Compound 4e showed high similarity with the control BMS compound (4e), with a Tanimoto similarity score of 0.8018. However, the model identified it as an ineffective molecule, and actual experiments showed that it was a poor PD1 / PD-L1 inhibitor with an IC50 of 1261 nM. On the other hand, the model identified compound 4b as a highly effective PD1 / PD-L1 inhibitor, and HTRF assay confirmed that it had a very good IC of 339.9 nM. 50 . However, its pairwise Tanimoto similarity score with the control compound (4a) was only 0.5074. In summary, these results indicate that the bootstrapped EGNN model can be used to select molecules for synthesis and experimental validation of PD-1 / PD-L1 inhibition, and it can identify inefficient molecules that are structurally similar to the control compound 4a.
[0210] Table 2. IC values of predicted active and inactive compounds 50 Value and EGNN SoftMax score
[0211]
[0212] *Indicates BMS active compounds.
[0213] Discussion and Conclusion
[0214] Cancer immunotherapy marks an important step in the treatment of cancer, and the development of PD-1 / PD-L1 immune checkpoint inhibitors has become an important research area for the treatment of many tumors. Currently, six therapeutic antibodies targeting PD-1 (pembrolizumab, nivolumab, and cemiplimab) and PD-L1 (atezolizumab, durvalumab, and avelumab) have been approved by the US FDA. In recent years, several new small molecule PD-1 / PD-L1 inhibitors have been developed. 43 , and structure determination of co-crystals of human PD-1 / PD-L1 complex and inhibitory ligands 44-46The field remains very active in the search for new small molecules to inhibit this important checkpoint, and we hope to increase the pace of this search by using new structure-based ML methods that have been extensively benchmarked and prospectively tested.
[0215] We have developed a new ML method, EGNN, to select, synthesize, and experimentally validate potent inhibitors of the PD-1 / PD-L1 interaction based on combining local features of small molecule topology and global features of small molecule interactions within the binding pocket as energy scores. Specifically, EGNN outperforms traditional ML architectures such as RF, SVM that include both local and global features, and GNN models that use only local features of small molecule topology. When benchmarked against known PD-1 / PD-L1 inhibitors from BMS and Incyte patent data, we conclude that the topology of the small molecule, the structural interactions in the binding pocket, and the chemical diversity of the training data are all important considerations for enhancing model performance.
[0216] We used bootstrapped EGNN models (based on 1000 EGNN models) to predict and confidently select novel molecules for chemical synthesis and subsequent inhibition testing using the HTRF PD-1 / PD-L1 inhibition assay. We believe that bootstrapping is an important statistical technique to use with ML methods in drug design to confidently select molecules for experimental validation. The predicted high potency molecule (4-((3-(2,3-dihydrobenzo[ b [1,4]dioxin-6-yl)-2-methylbenzyl)oxy)-2,6-dimethoxybenzyl)-D-serine is a hybrid of two BMS efficient molecular scaffolds and has an IC of 339.9 nM for inhibiting PD-1 / PD-L1 interaction 50 This indicates the ability of the EGNN model to perform backbone hopping to identify novel inhibitors. The accurate selection of inefficient molecules with different backbones demonstrates the practical utility of our bootstrap model in selecting compounds for synthesis, which is a difficult problem in the field of ML-based drug design.
[0217] Our EGNN approach can be further developed by adding more chemically diverse data and combining reinforced iterative learning with experiments performed at each step to develop a library of structurally diverse small molecules that inhibit PD-1 / PD-L1 interactions to guide structure-activity relationships. Given the general nature of the machine learning model and the easy-to-use docking method, this approach can be applied to identify small molecule immunomodulators by targeting other immune checkpoints, as well as generally for target-based drug design including local and global features.
[0218] Experimental Section
[0219] Homogeneous Time-Resolved Fluorescence (HTRF) Assay for Experimental Prediction of Inhibition of Compounds
[0220] Four highly potent and low potency predicted compounds were tested for inhibition of PD-1 / PD-L1 interaction using the PD1 / PD-L1 HTRF assay kit from Cisbio US, Inc. The assay protocol was used as described in the kit for each predicted compound (4b, 4c, 4d, and 4e) and the BMS control compound (4a). Briefly, 2 µL of compound, 4 µL of 25 nM Tag1-PD-L1 protein solution, and 4 µL of 250 nM Tag2-PD1 protein were added to Cisbio's HTRF 96-well low volume white plate. The plate was then incubated at room temperature for 15 minutes. Next, 10 µL of premixed anti-tag detection reagent (5 µL of 1X anti-Tag1-Eu 3+ and 5 µL of 1X anti-Tag2-XL665), and the sealed plate was incubated at room temperature for 2 hours. Finally, the plate seal was removed and HTRF ® The protocol used 12 different concentrations of each compound, with the maximum and minimum assay concentrations being 10,000 nM and 0.001 nM, respectively. Several replicates were run for the highly potent predicted compound 4b (36 data points) and the positive control compound 4a (48 data points). The fitted curves of the normalized signal expressed relative to log[concentration] (calculated using the HTRF ratio 665 nm / 620 nm) were used to determine the 50% inhibitory concentration (IC 50 ) (See next section IC 50 value calculation).
[0221] To calculate , first calculate the HTRF ratio as follows:
[0222] .
[0223] The multiplication factor of 10000 was not used to handle decimal values to improve data accuracy during calculations. The ΔR ratios indicating a "specific signal" for compounds disrupting the PD-1 / PD-L1 interaction were calculated by subtracting the background HTRF ratio (negative DMSO control in our work) from each compound (sample) HTRF ratio as follows:
[0224] .
[0225] Next, data normalization was performed to minimize variation in values when assays were performed on different days, on different plate reader instruments, or by different individuals. Normalization was performed relative to the background HTRF ratio and was calculated as follows:
[0226] .
[0227] Finally, calculate ratio to enable comparison of values between multiple experiments.
[0228]
[0229] Among them As a positive DMSO control in the assay .
[0230] IC 50 Calculation of values
[0231] The IC of PD-1 / PD-L1 inhibition was determined by analyzing the logarithm of the concentration-response curves to fit a sigmoidal curve with four-parameter logistic (4PL) regression using GraphPad Prism software version 8.3.0 for Windows (GraphPad Software, LaJolla California USA, www.graphpad.com). 50 Value. IC 50 The values are provided in Table 2. The following equation defines the regression curve.
[0232]
[0233] in = logarithm of concentration, = Answer , and = platform in the same units as Y, = same logarithmic units as X, = Slope factor or Hill slope, unitless. Using the above equation, calculate To obtain the IC of each compound 50 The HTRF_IC50_Data.xlsx data file containing all replicates is provided as a supporting file and was used in GraphPad Prism software to calculate the IC 50 value.
[0234] Machine learning architecture of the EGNN model
[0235] Using PyTorch 47Developing the EGNN model. All scripts used to implement the machine learning models and results are available on GitHub at https: / / github.com / chopralab / egnn. Figure 1 shows an overview of the EGNN machine learning architecture. We used the molecular graph of Tsubaki and colleagues 24 Implemented a graph neural network. In short, using ChemAxonMolConverter 48 The software converts the molecular structure into a SMILES string. Then use RDKit 49 Software package and Weisfeiler-Lehman algorithm to extract r-radius subgraphs of molecules ( Figure 1 A). The following sections include the details of the EGNN architecture.
[0236] Graph Neural Networks for Molecular Graphs in EGNN
[0237] The following equations and symbols along with the details of the molecular GNN have been adapted from the original work here 24 Copied from , with minor modifications for clarification. Lowercase bold letters (e.g. ) denotes a vector, and uppercase bold letters (e.g. ) indicates a matrix, and italic non-bold letters (e.g. and ) represents scalars, sets, graphs, vertices, and edges. GNN converts molecular graphs into low-dimensional real-valued vectors with two neural network-based functions ; Conversion and output 21 In Figure In each vertex The vertices are updated by taking into account the information of their neighboring vertices and edges through the transformation function. These vertices have been mapped into real-valued vectors through the output function . Both functions are differentiable. All the input features and weights of the GNN model are updated using back propagation with the help of the cross entropy loss function.
[0238] A graph can be defined as , here; and are the sets of vertices and edges, respectively. When applied to chemistry, atoms can be defined as vertices and chemical bonds can be defined as edges. First, all atoms and chemical bonds will be embedded as real-valued vectors with d dimensions based on their different types. Since the diversity of atoms (e.g., C, N, O, etc.) and bonds (e.g., single bonds, double bonds, triple bonds, etc.) in small molecules is limited, the number of learning parameters is limited. Therefore, a method called r-radius subgraph is used. 50 strategies to avoid this limitation.
[0239] r-radius subplot
[0240] In the definition The set of all atoms within the radius, an atom It can be expressed as .when , it is the collection of all the atoms in the molecule. Vertices The r-radius subgraph of is defined as follows;
[0241] ,in,
[0242]
[0243]
[0244] No. and Atoms The r-radius subgraph of the edges between is defined as follows;
[0245]
[0246] According to the type, randomly initialized embedding ( Figure 1 ) is assigned to each r-radius edge and vertices . Back-propagation has been used to train these random embeddings.
[0247] Vertex conversion function
[0248] explain is the time step Given a molecular graph No. The embedding vector of each vertex. Then update The vector can be written as follows;
[0249]
[0250] in, represents the set of adjacent atoms, is defined as The S-shaped function of is the hidden vector defining the neighborhood and can be calculated as follows;
[0251]
[0252] in, is the ReLU, a nonlinear activation function such as . and are the weight matrix and bias vector respectively. The following molecular graphs and The vector between atoms (vertices) is defined as .
[0253] Edge conversion function
[0254] As mentioned before, during training the edge vector of each embedding is updated using the edge transformation function .
[0255]
[0256]
[0257] in, , are the weight matrix and bias vector respectively. In addition, we add and , because the edges in the molecular graph have no direction.
[0258] Molecular vector output of molecular GNN
[0259] The transformation function generates a set of updated atomic (vertex) vectors The output function then uses this set of atomic vectors to obtain the unique molecule vector ( Figure 1 A), which is defined as follows:
[0260]
[0261] where the total number of vertices in the complete molecular graph is given by express.
[0262] Using docking and energy vector (E) to generate energy features in EGNN
[0263] First, use MarvinSketch 51 The software carefully draws all reported molecules. All drawn molecules are then cleaned in 3D and converted into sybyl.mol2 files which are used to interface with our in-house CANDOCK 31 package (version 0.6.0) with default parameters (with a max_num_possible of 20,000) and a 20% top seed percentage ( Figure 1B). CANDOCK source code is available on GitHub at https: / / github.com / chopralab / candock / releases / tag / v0.6.0. Docking was done using the PD-L1 homodimer crystal structure (PDB ID: 5N2F). We selected the binding site based on the coordinates of the crystal ligand in the protein structure (ligand ID: 8HW). Then, radial-mean-reduced-6 (RMR6) was added. 31 Used as a "Selector" parameter for docking to select the top pose 31 Next, the top pose of each docked compound was selected and CANDOCK 31 All 96 different potential energy functions available in the software were recalculated for their docking scores. All 96 CANDOCK docking energy scores for each molecule were normalized for each potential energy function to be used as a vector in the EGNN model;
[0264]
[0265] in, and Here, is the number of potential energy scoring functions, and is the number of molecules in the data set. It is The use of docking molecules The normalized docking energy value of the energy score of the potential energy function. Similarly, is the docking energy score before normalization. And, and is the first of all docked molecules The maximum and minimum energy values within the scoring function are then used using the scikit-learn package 46 The Cohen_κ_score tool in calcu- lates Cohen's κ scores for each scoring function for all training set data. All scoring functions that gave positive Cohen's κ scores were selected and the top one in each class was chosen for the EGNN model. Therefore, the normalized docking score vector for each molecule in the EGNN model using the potential energy scoring function normalized by RCR15 and RCC15 is expressed as ( Figure 1 B).
[0266] Output of EGNN
[0267] like Figure 1 As shown in C, the normalized docking energy score vector ( ) and the molecular vector output of GNN ( ) connection. Then, the long vector connected as follows Used for training to get the output vector ;
[0268]
[0269] in Indicates connection, represents the weight matrix, and Represents the bias vector. Then, the SoftMax classifier ( Figure 1 D) Add to The top of the vector to get high or low efficiency probability.
[0270]
[0271] in, ; 0 indicates inefficient and 1 indicates efficient, and is given probability.
[0272] Bootstrapping EGNN Model
[0273] The model uses random numbers to initialize edge and vertex vectors. Therefore, the final model uses bootstrapping to obtain predictions. One thousand different models with different random seeds were trained and predictions were obtained for the internal molecular design test set. The average SoftMax score was used as the final prediction result of the bootstrapped model. Finally, the molecules in the synthetic test set were classified as efficient or inefficient based on the average SoftMax score. If it is greater than or equal to 0.5, it is considered efficient, otherwise it is considered inefficient ( Figure 1 ). Therefore, the EGNN model uses the given SMILES string, which is generated by CANDOCK 31 The vectors of RCR15 and RCC15 scores generated and their correlation with the high or low efficiency status of PD-L1 protein are back-propagated for training. The trained model can be used to predict the probability that a given molecule is a high or low efficiency molecule for PD-L1 protein.
[0274] EGNN training and hyperparameter optimization
[0275] The model takes as input the SMILES string of a given molecule and the docking energy score string. The hyperparameters of the model were optimized before being used for predictions. The dimension (dim) of the GNN hidden vector, the number of hidden layers of the GNN, and the subgraph radius were optimized by considering the F1 score of the five-fold cross validation. Three values (i.e., dim = 5, 10, and 15) were used for the dimension of the GNN hidden molecule vector output. The numbers 1, 2, and 3 were used to check the optimal number of hidden layers in the GNN. Finally, the optimal subgraph radius of the model was selected from radius = 1, 2, and 3.
[0276] Calculation of F1 score and Cohen's κ
[0277] The following terms are used to calculate Cohen's κ and F1 scores. The number of compounds predicted to be highly effective and reported experimentally as highly effective is considered to be true positives (TP). The number of compounds predicted to be highly effective but reported experimentally as ineffective in the patent is considered to be false positives (FP). True negatives (TN) are defined as the number of compounds predicted to be ineffective and also reported experimentally as ineffective. False negatives (FN) are then defined as the number of compounds predicted to be ineffective but reported experimentally as highly effective. The F1 score is defined as follows.
[0278]
[0279] where precision and recall are defined as follows.
[0280]
[0281]
[0282] Cohen's κ:
[0283]
[0284] = Relative observed agreement between raters
[0285]
[0286] = Probability of random agreement
[0287] .
[0288] synthesis
[0289] Unless otherwise noted, all reagents and solvents were purchased from commercial sources and used as received. All reactions were performed in screw cap vials. Protons were acquired using 500 MHz using Me4Si as an internal standard ( 1 H) and carbon ( 13C) NMR spectra and are reported in δ units. Coupling constants are reported in Hz ( J Values). Column chromatography was performed on silica gel using flash chromatography (Teledyne ISCO EZprep). High resolution mass spectra (HRMS) were obtained using electrospray ionization (ESI) technology and a TOF mass analyzer. Organic solvents and starting materials were used as received. BMS compound 4a (BMS-1 or KPGC01S94) was synthesized from compounds 1, 2a-b, 3a-b according to the reported procedure 7 and compounds 4b-c, and the spectral data are consistent with the reported data 6-8 .
[0290] Compound 4a (BMS-1 or KPGC01S94), ( S )-1-(2,6-dimethoxy-4-((2-methyl-[1,1'-biphenyl]-3-yl)methoxy)benzyl)piperidine-2-carboxylic acid: 3b (45 mg, 0.125 mmol) from Scheme 1, (S)-piperidine-2-carboxylic acid (64.5 mg, 4 eq, 0.5 mmol), sodium cyanoborohydride (40.8 mg, 5.2 eq, 0.65 mmol) were dissolved in DMF (1 mL) and acetic acid (2 drops) was added. The reaction mixture was stirred at 80 °C for 1 hour. The reaction was monitored by TLC. The crude was purified by 0-20% DCM:MeOH to afford the desired product (31.5 mg, 53% yield) as an off-white solid. 1 H NMR (500 MHz, DMSO- d 6 ) δ 7.49 - 7.41 (m, 3H), 7.39 - 7.34 (m, 1H), 7.32 - 7.25 (m, 3H), 7.19 (dd, J = 7.7, 1.5 Hz, 1H), 6.41 (s, 2H), 5.17 (s,2H), 4.08 (s, 2H), 3.78 (s, 7H), 3.11 (t, J = 5.5, 5.5 Hz, 1H), 3.08 - 2.99(m, 1H), 2.60 (dd, J = 13.5, 6.7 Hz, 1H), 2.20 (s, 3H), 1.80 (q, J = 6.0,5.9, 5.9 Hz, 2H), 1.55 (q, J= 6.7, 6.1, 6.1 Hz, 2H), 1.37 (ddt, J = 18.4,12.8, 6.5, 6.5 Hz, 3H); 13 C NMR (126 MHz, DMSO- d 6 ) δ 172.0, 161.47, 160.32,142.70, 141.85, 135.83, 134.59, 130.25, 129.63, 128.86, 128.72, 127.44,126.04, 92.05, 69.23, 64.35, 56.42, 48.70, 46.21, 31.16, 26.11, 22.13, 21.27,16.41.
[0291] Compound 4b (KPGC01S32), (4-((3-(2,3-dihydrobenzo[ b [1,4]dioxin-6-yl)-2-methylbenzyl)oxy)-2,6-dimethoxybenzyl)-D-serine: 3a (35.6 mg, 0.104 mmol) from Scheme 1, D-serine (32.8 mg, 3 eq.), sodium cyanoborohydride (19.6 mg, 3 eq.) were dissolved in DMF (1 mL) and acetic acid (0.104 mmol, 1 eq., 2 drops) was added. The reaction mixture was stirred at room temperature overnight. The reaction was monitored by TLC. The crude was purified by 0-20% MeOH:DCM to afford the desired product (42% yield) as an off-white solid. 1 HNMR (500 MHz, DMSO- d6) δ 7.42 (dd, J = 7.6, 1.5 Hz, 1H), 7.22 (t, J = 7.6,7.6 Hz, 1H), 7.15 (dd, J = 7.6, 1.5 Hz, 1H), 6.90 (d, J = 8.2 Hz, 1H), 6.76(d, J = 2.1 Hz, 1H), 6.73 (dd, J = 8.2, 2.1 Hz, 1H), 6.37 (s, 2H), 5.13 (s,2H), 4.26 (s, 4H), 3.86 (s, 2H), 3.77 (s, 6H), 3.58 (dt, J = 8.5, 3.4, 3.4Hz, 2H), 2.94 (t, J = 6.0, 6.0 Hz, 1H, NH), 2.20 (s, 3H); 13 C NMR (126 MHz, DMSO- d 6) δ 172.85, 160.75, 159.69, 159.53, 143.42, 142.96, 142.12, 135.85,134.96, 134.65, 130.20, 128.56, 125.93, 122.59, 118.17, 117.26, 104.65,92.06, 69.18, 64.57, 62.51, 61.34, 56.43, 56.32, 16.41; HRMS (ESI): for C 28 H 32 NO8[M + H] + , measured 510.2132 m / z; calculated mass, 510.2128.
[0292] Compound 4c (KPGC01S138), N -(2,6-Dimethoxy-4-((2-methyl-[1,1'-biphenyl]-3-yl)methoxy)benzyl)-3,3,3-trifluoro-1-phenylpropane-1-amine: 3b (8 mg, 0.022 mmol) from Scheme 1, 3,3,3-trifluoro-1-phenylpropane-1-amine (16.7 mg, 0.088 mmol, 4 eq.), sodium cyanoborohydride (7.2 mg, 0.114 mmol, 5.2 eq.) were dissolved in DMF (0.5 mL) and acetic acid (1 drop) was added. The reaction mixture was stirred at 80 °C for 3 hours. The reaction was monitored by TLC. The crude was purified by 0-20% MeOH:DCM to afford the desired product (68% yield) as an oily product.1 H NMR (500 MHz, CDCl3) δ 7.45 - 7.38 (m, 5H), 7.38 - 7.30 (m,6H), 7.30 - 7.27 (m, 1H), 7.26 (d, J = 5.4 Hz, 1H), 6.22 (s, 2H), 5.08 (s,2H), 3.96 (t, J = 6.5, 6.5 Hz, 1H), 3.75 (s, 6H), 3.73 (d, J = 6.5 Hz, 2H), 2.46 (s, 1H), 2.27 (s, 3H), 1.29 - 1.24 (m, 1H); 13 C NMR (126 MHz, CDCl3) δ160.23, 159.41, 143.06, 141.94, 135.1, 134.49, 132.81, 130.34, 129.40,128.49, 128.32, 128.11, 127.91, 127.62, 127.18, 126.90, 125.65, 91.17, 69.34,56.21, 55.53, 39.11, 16.23; HRMS (ESI): for C 32 H 33 F3NO3[M + H] + , measured 536.2419 m / z; calculated mass, 536.2413.
[0293] Compound 4d (GCL.2), ( 7R,8R,9S,13S,14S,17R )-17-ethynyl-17-hydroxy-7,13-dimethyl-1,2,6,7,8,9,10,11,12,13,14,15,16,17-tetradecahydro-3 H -Cyclopenta[a]phenanthrene-3-one: Tibolone (156 mg, 0.5 mmol) was placed in a round bottom flask containing 10 mL of THF and 100 µL of water was added. Next, p-toluenesulfonic acid (85 mg, 0.5 mmol) was added to it and the mixture was refluxed at 80 °C for 48 hours and the progress of the reaction was monitored by TLC. The organic solvent was then evaporated to dryness to get the crude product, which was purified by flash column chromatography using 20% ethyl acetate solvent mixture in petroleum ether as eluent to produce pure compound GCL2 as an off-white solid (53% yield). 1H NMR (500 MHz, MeOD) δ 5.80 (t, J = 2.1, 2.1 Hz, 1H), 2.88 (s, 1H), 2.56 (ddt, J = 14.1, 5.4, 1.6, 1.6 Hz, 1H), 2.42 - 2.28 (m, 4H), 2.27 - 2.19 (m,1H), 2.18 - 2.13 (m, 1H), 2.06 - 1.90 (m, 3H), 1.77 - 1.53 (m, 6H), 1.44 -1.25 (m, 2H), 1.14 (qd, J = 11.0, 11.0, 10.9, 4.2 Hz, 1H), 0.91 (d, J = 0.7Hz, 3H), 0.79 (d, J = 7.1 Hz, 3H). 13 C NMR (126 MHz, MeOD) δ 201.06, 167.84,125.32, 87.28, 78.82, 73.45, 48.12, 47.95, 47.78, 47.61, 47.44, 47.27, 47.10,46.67, 45.82, 43.05, 42.76, 42.18, 38.28, 36.02, 32.25, 30.66, 26.44, 26.41,21.71, 11.79, 11.77.
[0294] Compound 4e (KPGC01S42), ( S )-1-(2,6-dimethoxy-4-((2-methyl-[1,1'-biphenyl]-3-yl)methoxy)benzyl)piperidine-3-carboxylic acid: 3b (24 mg, 0.0066 mmol) from Scheme 1, (D)-3-piperidinecarboxylic acid (34.2 mg, 0.0265 mmol, 4 eq.), sodium cyanoborohydride (5.2 mg, 0.0343 mmol, 5.2 eq.) were dissolved in DMF (1 mL) and acetic acid (1 drop) was added. The reaction mixture was stirred at room temperature for 14 hours. The reaction was monitored by TLC (silica gel, 5% DCM:MeOH). The crude material was purified by flash chromatography using 0-20% DCM:MeOH to provide the desired product (43% yield) as an oily product. 1 H NMR (500 MHz, DMSO): δ 7.49 - 7.41 (m, 3H), 7.39- 7.34 (m, 1H), 7.32 - 7.25 (m, 3H), 7.19 (d,J = 7.7 Hz, 1H), 6.37 (s, 2H), 5.15 (s, 2H), 3.75 (s, 6H), 3.53 (s, 2H), 2.38 - 2.30 (m, 2H), 2.20 (s, 3H), 1.97 (s, 4H), 1.16 (s, 2H); 13 C NMR (126 MHz, DMSO): δ 170.81, 160.50, 160.06,142.67, 141.87, 135.99, 134.56, 130.18, 129.63, 128.81, 128.71, 127.42,126.01, 92.00, 69.09, 60.22, 56.27, 55.37, 55.04, 52.65, 49.05, 31.14, 26.68,21.21, 16.40, 14.54; LCMS / MS (ESI): for C 29 H 33 NO5[M + H] + , measured 476.3 m / z; calculated mass, 476.24.
[0295]
[0296] Synthesis of 2a:
[0297] Step-1: 2-(3-Bromo-2-methylphenyl)-5-((4-fluorophenyl) (piperidin-1-yl)methyl)-1,3,4-oxadiazole: In a clean, dry screw cap vial with a magnetic stir bar, a mixture of piperidine (1 eq.), 4-fluorobenzaldehyde (1 eq.), N-(isocyanatoimido)triphenylphosphine (1 eq.) was dissolved in DCM (5 mL / mmol). A solution of 3-bromo-2-methyl-benzoic acid (1 eq.) in DCM was slowly added to the reaction mixture at room temperature and stirred at 50-60°C for 2 hours. The solvent was removed in vacuo and the crude was purified by flash chromatography using hexane:ethyl acetate (0-60%) as eluent to obtain the product as a light yellow oil.
[0298] Step-2: 2-((4-Fluorophenyl)(piperidin-1-yl)methyl)-5-(2-methyl-[1,1'-biphenyl]-3-yl)-1,3,4-oxadiazole (2a): In a clean, dry screw cap vial with a magnetic stir bar, a mixture of 2-(3-bromo-2-methylphenyl)-5-((4-fluorophenyl)(piperidin-1-yl)methyl)-1,3,4-oxadiazole (1 eq.), phenylboronic acid (2 eq.) and PdCl2(dppf)2-CH2Cl2 (3 mol%) was placed and purged once with argon. Toluene (4.5 mL) and ethanol (1.5 mL) were added and the reaction mixture was purged with argon. While purging, 1.5 mL of 1M NaHCO3 was added and it was stirred at 80°C for 45 minutes. The progress of the reaction was monitored by TLC. After the completion of the reaction, ethyl acetate (20 mL) was added and washed with water (2 x 20 mL). The organic solvent was removed under reduced pressure and the crude was purified by flash chromatography (0-60%, hexanes:ethyl acetate) to give the product as a clear oil.
[0299] Synthesis of 2b
[0300] In a clean, dry screw cap vial with a magnetic stir bar, a mixture of 2-(3-bromo-2-methylphenyl)-5-((4-fluorophenyl)(piperidin-1-yl)methyl)-1,3,4-oxadiazole (1 eq), (2,3-dihydrobenzo[b][1,4]dioxin-6-yl)boronic acid (2 eq) and PdCl2(dppf)2-CH2Cl2 (3 mol%) was placed and purged once with argon. Toluene (4.5 mL) and ethanol (1.5 mL) were added and the reaction mixture was purged with argon. While purging, 1.5 mL of 1M NaHCO3 was added and stirred at 80°C for 45 minutes. The progress of the reaction was monitored by TLC. After the reaction was complete, ethyl acetate (20 mL) was added and washed with water (2 x 20 mL). The organic solvent was removed under reduced pressure and the crude was purified by flash chromatography (0-60%, hexane:ethyl acetate) to give the product as a clear oil.
[0301] Synthesis of 2c
[0302]
[0303] Preparation of N-(tert-butoxycarbonyl)-O-(tert-butyl)-D-serine methyl ester: A mixture of Boc-D-serine (1 equivalent), iodomethane (3 equivalents) in DMF was stirred at room temperature for 2 hours. Ice-cold water was added to the reaction mixture and extracted with ethyl acetate. The organic layer was then washed with water and the ethyl acetate was removed in vacuo. The crude oil was obtained quantitatively and used in the next step without further purification.
[0304] Preparation of tert-butyl (R)-(3-(tert-butoxy)-1-hydrazine-1-oxopropan-2-yl)carbamate: Hydrazine monohydrate (2 eq.) was slowly added to a mixture of methyl ester of Boc-D-serine (1 eq.) in DCM kept in an ice bath. It was stirred at room temperature for 2 hours until a white suspension was observed. Ice-cold water was added to the reaction mixture and extracted with ethyl acetate. The organic layer was then washed with water and the ethyl acetate was removed in vacuo. The product was obtained quantitatively as a white solid and was used in the next step without further purification.
[0305]
[0306] A mixture of 3-bromo-2-methylbenzoic acid (1 eq.), EDCI (1.5 eq.), DIPEA (5 eq.) in DMF was stirred for 10 minutes, and then (R)-tert-butyl (3-(tert-butoxy)-1-hydrazino-1-oxopropane-2-yl)carbamate (1 eq.) was added to the reaction mixture. It was stirred at room temperature overnight. The reaction was monitored by TLC. The crude material was extracted with ethyl acetate: ice-cold water, and the ethyl acetate layer was removed under reduced pressure. The product was purified by flash column chromatography using 0-60% hexane:ethyl acetate to obtain a white solid (65% yield).
[0307]
[0308] A mixture of (R)-(1-(2-(3-bromo-2-methylbenzoyl)hydrazine)-3-(tert-butoxy)-1-oxopropane-2-yl)carbamate (1 eq., 0.5 mmol), PPh (1.1 eq.), iodine (2 eq.), triethylamine (2 eq.) in THF (5 mL) was stirred overnight at room temperature. The reaction was monitored by TLC. The product was purified by flash column chromatography using 0-60% hexane:ethyl acetate to give an oily product (78% yield).
[0309]
[0310] (R)-tert-butyl (1-(5-(3-bromo-2-methylphenyl)-1,3,4-oxadiazol-2-yl)-2-(tert-butoxy)ethyl)carbamate (1 equivalent), (2,3-dihydrobenzo[ b A mixture of [1,4]dioxin-6-yl)boronic acid (1.5 equiv), PdCl2(dppf)•CH2Cl2 (3 mol%) in toluene:ethanol (1.5:0.5 mL) was purged with argon twice. 1M NaHCO3 (1.5 mL) was added under inert atmosphere, and it was stirred at 80°C for 45 minutes. The reaction was monitored by TLC. The product was purified by flash column chromatography using 0-60% hexane:ethyl acetate to give an oily product (63% yield).
[0311]
[0312] To a solution of (R)-tert-butyl(2-(tert-butoxy)-1-(5-(3-(2,3-dihydrobenzo[b][1,4]dioxin-6-yl)-2-methylphenyl)-1,3,4-oxadiazol-2-yl)ethyl)carbamate (50 mg) in DCM (2 mL) was added trifluoroacetic acid (0.5 mL) and stirred at room temperature for 3 hours. The reaction was monitored by TLC. The product was purified by flash column chromatography using 0-50% DCM: methanol to give the product as an oil (88% yield).
[0313]
[0314] Synthesis of (IIIa)
[0315]
[0316] Step-1: 3-Bromo- N ,2-dimethylaniline (1 equivalent), (2,3-dihydrobenzo[ b [1,4]dioxin-6-yl)boronic acid (1.5 equiv), PdCl2(dppf)•CH2Cl2 (3 mol%) were purged with argon twice. 1M NaHCO3 (1.5 mL) was added under inert atmosphere and it was stirred at 80°C for 45 minutes. The reaction was monitored by TLC. The product was purified by flash column chromatography using 0-60% hexane:ethyl acetate to give an oily product (77% yield).
[0317]
[0318] Step-2: To 3-(2,3-dihydrobenzo[ b[1,4]dioxin-6-yl)- N , a mixture of 2-dimethylaniline (1 eq.), propionaldehyde (1.1 eq.), (N-isocyanoimino)triphenylphosphine in DCM was added portionwise with benzoic acid (1.1 eq.). The reaction mixture was stirred at 40 °C for 2-3 hours. The reaction was monitored by TLC. The product was purified by flash column chromatography using 0-60% hexane:ethyl acetate to give an oily product (54% yield).
[0319] Synthesis of (IIIb)
[0320] Step-1: 3-Bromo- N ,2-Dimethylaniline (1 eq.), phenylboronic acid (1.5 eq.), PdCl2(dppf)•CH2Cl2 (3 mol%) were purged with argon twice. 1M NaHCO3 (1.5 mL) was added under inert atmosphere and it was stirred at 80°C for 45 minutes. The reaction was monitored by TLC. The product was purified by flash column chromatography using 0-60% hexane:ethyl acetate to give an oily product (74% yield).
[0321]
[0322] Step-2: To a mixture of N,2-dimethyl-[1,1'-biphenyl]-3-amine (1 eq.), propionaldehyde (1.1 eq.), (N-isocyanoimido)triphenylphosphine in DCM was added benzoic acid (1.1 eq.) portionwise. The reaction mixture was stirred at 40°C for 2-3 hours. The reaction was monitored by TLC. The product was purified by flash column chromatography using 0-60% hexane:ethyl acetate to obtain an oily product (48% yield).
[0323]
[0324]
[0325]
[0326] Representative Procedure: 4-(5-phenyl-1,3,4-oxadiazol-2-yl)-3,4-dihydrobenzo[e][1,2,3]oxathiazine 2,2-dioxide [derivative of Formula IV]:
[0327]
[0328] In a clean, oven-dried screw-cap vial, sulfonylimide 1a (0.2 mmol), benzoic acid (1.1 eq.), (N-isocyanoimino)triphenylphosphine (1.1 eq.) were placed and cooled to -10 °C. Next, solvent CH2Cl2 (3 mL) was added and stirred at the same temperature for 5-10 minutes. The solvent was partially removed under reduced pressure, and the crude material was directly loaded onto a silica gel cartridge and purified by flash chromatography (Teledyne ISCO) using EtOAc: Hexane as eluent. The main fractions were collected and repurified by reverse phase preparative HPLC (Teledyne ISCO) using water: MeCN as eluent.
[0329] Compounds having formula V:
[0330] A compound according to formula (V), wherein:
[0331] When R1 = CH3, the compound is
[0332]
[0333]
[0334] When R1 = Cl, the compound is
[0335]
[0336] For the synthesis of 4-methyl- N Representative procedure for -(phenyl(5-phenyl-1,3,4-oxadiazol-2-yl)methyl)benzenesulfonamide [Formula V derivative]: In a clean, oven-dried screw cap vial, N-tosyl imide (0.2 mmol), benzoic acid (1.1 eq), (N-isocyanoimido)triphenylphosphine (1.1 eq) were placed and cooled to -10°C. Next, solvent CHCl (3 mL) was added and stirred at the same temperature for 5-10 minutes. The solvent was partially removed under reduced pressure and the crude was directly loaded onto a silica gel cartridge and purified by flash chromatography (Teledyne ISCO) using EtOAc: Hexane as eluent to obtain a solid product.
[0337] Those skilled in the art will recognize that numerous modifications may be made to the above specific implementations. The implementations should not be limited to the specific limitations described. Other implementations are also possible.
[0338] Although the present invention has been described and illustrated in detail in the accompanying drawings and the foregoing description, it should be regarded as exemplary rather than restrictive in nature, and it should be understood that only certain embodiments are shown and described, and all changes and modifications that fall within the spirit of the present invention are desired to be protected. The scope of the method and apparatus of the present invention is intended to be limited by the following claims. However, it must be understood that the present disclosure may be implemented in a manner other than that specifically explained and described without departing from its spirit or scope.
[0339] References
[0340] (1) Pardoll, DM The Blockade of Immune Checkpoints in CancerImmunotherapy. Nat. Rev. Cancer 2012, 12 (4), 252-264.
[0341] (2) Buchbinder, EI; Desai, A. CTLA-4 and PD-1 PathwaysSimilarities, Differences, and Implications of Their Inhibition. American Journal of Clinical Oncology: Cancer Clinical Trials . Wolters Kluwer Health 2016, pp. 98-106.
[0342] (3) Sasikumar, PGN; Ramachandra, M.; Naremaddepalli, SSS 1,2, 4-Oxadiazole Derivatives as Immunomodulators. Google Patents September 2017.
[0343] (4) Sunshine, J.; Taube, JM Pd-1 / Pd-L1 Inhibitors. Curr. Opin. Pharmacol. 2015, 23 , 32-38.
[0344] (5) Imai, K.; Takaoka, A. Comparing Antibody and Small-MoleculeTherapies for Cancer. Nat. Rev. Cancer 2006, 6(9), 714 - 727。
[0345] (6) Chupak, L.; Ding, M.; Martin, S.; Connoly, T.; Xu, N.; Yeung, K.; Zhu, J.; Langley, D.; et al. Compounds Useful as Immunomodulators. WO / 2015 / 160641, October 22, 2015.
[0346] (7) Chupak, L. S.; Zheng, X.; Mingo, P. A. Compounds Useful as Immunomodulators. WO / 2015 / 034820, December 3, 2015.
[0347] (8) Guzik, K.; Zak, K. M.; Grudnik, P.; Magiera, K.; Musielak, B.; Törner, R.; Skalniak, L.; Dömling, A.; Dubin, G.; Holak, T. A. Small - Molecule Inhibitors of the Programmed Cell Death - 1 / Programmed Death - Ligand 1 (PD - 1 / PD - L1) Interaction via Transiently Induced Protein States and Dimerization of PD - L1. J. Med. Chem. 2017, 60 (13), 5857 - 5867。
[0348] (9) Zak, K. M.; Grudnik, P.; Guzik, K.; Zieba, B. J.; Musielak, B.; Dömling, A.; Dubin, G.; Holak, T. A. Structural Basis for Small Molecule Targeting of the Programmed Death Ligand 1 (PD - L1). Oncotarget 2016, 7 (21), 30323 - 30335。
[0349] (10) Broach, J. R.; Thorner, J. High-Throughput Screening for DrugDiscovery. Nature . [London: Macmillan Journals], 1869 - 1996, pp. 14 - 16.
[0350] (11) Artis, D. R.; Lin, J. J.; et al. Scaffold-Based Discovery ofIndeglitazar, a PPAR Pan-Active Anti-Diabetic Agent. Proc. Natl. Acad. Sci. USA 2009, 106 (1), 262 - 267.
[0351] (12) Böhm, H.-J.; Flohr, A.; Stahl, M. Scaffold Hopping. Drug Discov. today Technol. 2004, 1 (3), 217 - 224.
[0352] (13) Kitchen, D. B.; Decornez, H.; Furr, J. R.; Bajorath, J. Dockingand Scoring in Virtual Screening for Drug Discovery: Methods andApplications. Nat. Rev. Drug Discov. 2004, 3 (11), 935 - 949.
[0353] (14) Joachims, T. Text Categorization with Support Vector Machines:Learning with Many Relevant Features. Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics) 1998, 1398 , 137 - 142.
[0354] (15) Burbidge, R.; Trotter, M.; Holden, S. Drug Design by MachineLearning: Support Vector Machines for Pharmaceutical Data Analysis. Comput. Chem. 2001, 26 (1), 5-14。
[0355] (16) Heikamp, K.; Bajorath, J. Support Vector Machines for DrugDiscovery. Expert Opin. Drug Discov. 2014。
[0356] (17) Breiman, L. Random Forests. Mac h. Learn. 2001, 9 (1), 5-32。
[0357] (18) Li, H.; Leung, K. S.; Wong, M. H.; Ballester, P. J. ImprovingAutodock Vina Using Random Forest: The Growing Accuracy of Binding AffinityPrediction by the Effective Exploitation of Larger Data Sets. Mol. Inform. 2015, 34 (2-3), 115-126。
[0358] (19) Zhang, Q. Y.; Aires-de-Sousa, J. Random Forest Prediction ofMutagenicity from Empirical Physicochemical Descriptors. J. Chem. Inf. Model. 2007, 47 (1), 1-8。
[0359] (20) Torng, W.; Altman, R. B. Graph Convolutional Neural Networks forPredicting Drug-Target Interactions. J. Chem. Inf. Model. 2019。
[0360] (21) Scarselli, F.; Gori, M.; Tsoi, A. C.; Hagenbuchner, M.;Monfardini, G. The Graph Neural Network Model. IEEE Trans. Neural Networks 2009, 20 (1), 61-80。
[0361] (22) Lim, J.; Ryu, S.; Park, K.; Choe, Y. J.; Ham, J.; Kim, W. Y.Predicting Drug-Target Interaction Using a Novel Graph Neural Network with 3DStructure-Embedded Graph Representation. J. Chem. Inf. Model. 2019, 59 (9),3981-3988。
[0362] (23) Ma, T.; Zhang, A. AffinityNet: Semi-Supervised Few-Shot Learning for Disease Type Prediction 。
[0363] (24) Tsubaki, M.; Tomii, K.; Sese, J. Compound-Protein InteractionPrediction with End-to-End Learning of Neural Networks for Graphs andSequences. Bioinformatics 2019, 35 (2), 309-318。
[0364] (25) Gupta, A.; Müller, A. T.; Huisman, B. J. H.; Fuchs, J. A.;Schneider, P.; Schneider, G. Generative Recurrent Networks for De Novo DrugDesign. Mol. Inform. 2018, 37 (1)。
[0365] (26) Ståhl, N.; Falkman, G.; Boström, J. Deep Reinforcement Learningfor Multiparameter Optimization in de Novo Drug Design. J. Chem. Inf. Model. 2019。
[0366] (27) Li, Y.; Zhang, L.; Liu, Z. Multi-Objective de Novo Drug Designwith Conditional Graph Generative Model. J. Cheminform. 2018, 10 (1)。
[0367] (28) Esaki, T.; Ohashi, R.; Watanabe, R.; Natsume-Kitatani, Y.;Kawashima, H.; Nagao, C.; Mizuguchi, K. Computational Model to Predict theFraction of Unbound Drug in the Brain. J. Chem. Inf. Model. 2019。
[0368] (29) Fine, J.; Chopra, G. Lemon: A Framework for Rapidly MiningStructural Information from the Protein Data Bank. Bioinformatics 2019。
[0369] (30) Chopra, G.; Samudrala, R. Exploring Polypharmacology in DrugDiscovery and Repurposing Using the CANDO Platform. Curr. Pharm. Des. 2016, 22 (21), 3109-3123。
[0370] (31) Fine, J.; Konc, J.; Samudrala, R.; Chopra, G. CANDOCK: ChemicalAtomic Network Based Hierarchical Flexible Docking Algorithm. bioRxiv 2018。
[0371] (32) Kischuk, E.; Majumder, J.; Fine, J. A.; Lantz, T. C.; Dhawan,D.; Knapp, D. W.; Ratliff, T. L.; Chopra, G. Abstract 4709: Cell-SpecificGene Program-Based Small-Molecule Immunomodulators Targeting Solid-TumorMicroenvironments; 2018。
[0372] (33) Esensten, J. H.; Helou, Y. A.; Chopra, G.; Weiss, A.; Bluestone,J. A. CD28 Costimulation: From Mechanism to Therapy. Immunity . 2016, pp. 973-988。
[0373] (34) Hernandez-Perez, M.; Chopra, G.; Fine, J.; Conteh, A. M.; Anderson, R. M.; Linnemann, A. K.; Benjamin, C.; Nelson, J. B.; Benninger, K. S.; Nadler, J. L.; et al. Inhibition of 12 / 15-Lipoxygenase Protects Against β-Cell Oxidative Stress and Glycemic Deterioration in Mouse Models of Type 1 Diabetes. Diabetes 2017, 66 (11), 2875-2887。
[0374] (35) Ma, X.; Zhou, J.; Wang, C.; Carter-Cooper, B.; Yang, F.; Larocque, E.; Fine, J.; Tsuji, G.; Chopra, G.; Lapidus, R. G.; et al. Identification of New FLT3 Inhibitors Potently Inhibit AML Cell Lines via an Azo Click-It / Staple-It Approach。 ACS Med. Chem. Lett. 2017, 8 (5)。
[0375] (36) Fine, J.; Lackner, R.; Samudrala, R.; Chopra, G. Computational Chemoproteomics to Understand the Role of Selected Psychoactives. Sci. Rep. 2019。
[0376] (37) Head, L.; Zhang, F.; Li, J.; Wang, H.; Xiao, K.; Wu, L.; Qian, D.-Q.; Lajkiewicz , N. ; Concol, L.; Li, Z.Curr. Heterocyclic CompoundsDerivatives as PD-L1 Internalization Inducers. WO / 2018 / 119263, 28 June 2018.
[0377] (38) Xiao, K.; Zhang, F.; Wu, L.; Yao, W. Heterocyclic Compounds asImmunomodulators. US 2018 / 0273519 A1, September 27, 2018.
[0378] (39) Willett, P. Similarity-Based Virtual Screening Using 2DFingerprints. Drug Discov. Today 2006, 11 (23-24), 1046-1
[0379] (40) DeLano, WL Pymol: An Open-Source Molecular Graphics Tool ;CCP4, 2002;No.
[0380] (41) Max, K.; Weston, S.; Keefer, C.; Engelhardt , A. ; Cooper , T. ;Mayer , Z. ; Kenkel, B.; Team, RC; Benesty , M. ; Lescarbeau, R.Classification and Regression Training. 2016,198 views.
[0381] (42) Robin, X.; Turck, N.; Hainard , A. ; Tiberti, N.; Lisacek , F. ;Sanchez , JC ; Müller, M. PROC: An Open-Source Package for R and S+ toAnalyze and Compare ROC Curves. BMC Bioinformatics 2011,12 (1), 77。
[0382] (43) Guzik, K.; Tomala, M.; Muszak, D.; Konieczny, M.; Hec, A.; Błaszkiewicz, U.; Pustuła, M.; Butera, R.; Dömling, A.; Holak, T. A.Development of the Inhibitors That Target the PD-1 / PD-L1 Interaction—A BriefLook at Progress on Small Molecules, Peptides and Macrocycles. Molecules 2019, 24 (11), 2071。
[0383] (44) Zak, K. M.; Kitel, R.; Przetocka, S.; Golik, P.; Guzik, K.;Musielak, B.; Dömling, A.; Dubin, G.; Holak, T. A. Structure of the Complexof Human Programmed Death 1, PD-1, and Its Ligand PD-L1. Structure 2015。
[0384] (45) Lee, H. T.; Lee, S. H.; Heo, Y. S. Molecular Interactions ofAntibody Drugs Targeting PD-1, PD-L1, and CTLA-4 in Immuno-Oncology. Molecules . 2019。
[0385] (46) Zak, K. M.; Grudnik, P.; Magiera, K.; Holak, T. A. StructuralBiology of the Immune Checkpoint Receptor PD-1 and Its Ligands PD-L1 / PD-L2. Structure . 2017。
[0386] (47) PyTorch Community. Tensors and Dynamic Neural Networks in Python with Strong GPU Acceleration. Github 2016, 6 。
[0387] (48) Chemaxon, M. Molecule File Converter, V. 5.10. 1,(C) 1999−2012 ChemAxon Ltd。
[0388] (49) Landrum, G. A. RDKit: Open-Source Cheminformatics。
[0389] (50) Costa, F.; De Grave, K. Fast Neighborhood Subgraph Pairwise Distance Kernel. In ICML 2010 - Proceedings, 27th International Conference on Machine Learning ; Omnipress; Madison, WI, USA, 2010; pp 255-262。
[0390] (51) Csizmadia, P. MarvinSketch and MarvinView: Molecule Applets for the World Wide Web. In Proceedings of ECSOC-3, the third international electronic conference on synthetic organic chemistry, September 1 - 30, 2019; page 1775。
Claims
1. A compound having formula (I) or a pharmaceutically acceptable salt thereof: in Ar1 is 2,3-dihydrobenzo[b][1,4]-dioxinyl; R1 and R2 are independently methoxy; R3 is C 1-6 alkyl; and Q is:
2. The compound according to claim 1 or a pharmaceutically acceptable salt thereof, wherein R3 is CH3.
3. The compound according to claim 1 or a pharmaceutically acceptable salt thereof, wherein the compound is 4. A pharmaceutical composition comprising one or more compounds according to any one of claims 1 to 3 or pharmaceutically acceptable salts thereof and one or more excipients or carriers. 5 . The compound or pharmaceutically acceptable salt thereof according to any one of claims 1 to 3 , wherein the compound or pharmaceutically acceptable salt thereof is an immunomodulator. 6 . The compound or pharmaceutically acceptable salt thereof according to any one of claims 1 to 3 , wherein the compound or pharmaceutically acceptable salt thereof is an inhibitor of PD-1 and PDL-1 signaling pathways.
7. The compound or pharmaceutically acceptable salt thereof according to any one of claims 1 to 3, wherein the compound or pharmaceutically acceptable salt thereof is used for treating cancer.
8. Use of a pharmaceutical composition in the preparation of a medicament for treating cancer, the pharmaceutical composition comprising one or more compounds according to any one of claims 1 to 3 or pharmaceutically acceptable salts thereof and one or more carriers or excipients.
9. Use of a pharmaceutical composition in the preparation of a medicament for treating cancer, the pharmaceutical composition comprising a compound according to any one of claims 1 to 3 or a pharmaceutically acceptable salt thereof in combination with one or more other compounds having the same or different mode of action and one or more carriers or excipients.
10. The use according to claim 9, wherein the cancer is castration-resistant prostate cancer.
11. A pharmaceutical composition for use as a cancer medicament, comprising one or more compounds according to any one of claims 1 to 3 or pharmaceutically acceptable salts thereof and one or more excipients or carriers.
