A method for screening of cdk5 and pbk inhibitors in pituitary neuroendocrine tumor based on artificial intelligence

Abstract

The application discloses a kind of CDK5 and PBK inhibitor drug screening method in pituitary neuroendocrine tumor based on artificial intelligence, obtains CDK5 / PBK selected inhibitor dataset from ChEMBL database;A number of prediction models are established using machine learning algorithm and deep learning framework coupled with molecular fingerprint and molecular graph, the hyperparameters of each prediction model are trained and optimized by training set and validation set, each prediction model is evaluated by test set;Select the prediction model as inhibitor drug screening model;The molecular inhibitory activity from FDA approved drug database and natural compound database is predicted using inhibitor drug screening model;Molecules with a prediction confidence score greater than 0.6 for both CDK5 and PBK are selected for evaluation, and CDK5 and PBK inhibitor drugs are obtained.The present application obtains antitumor drug 5 as a dual-target inhibitor through artificial intelligence-driven drug screening, and the development of dual-target inhibitor is specifically aimed at the key limitation of compensatory kinase activation observed in traditional single-target therapy.

Description

Technical Field

[0001] This invention belongs to the field of medical technology, specifically relating to an artificial intelligence-based method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors. Background Technology

[0002] Pituitary neuroendocrine tumors (PitNETs) are the second most common type of intracranial tumor, accounting for 17.4% of all primary central nervous system tumors. 35%–60% of these cases are invasive, infiltrating the dura mater, bone, or adjacent structures such as the cavernous sinus and sphenoid sinus. These invasive pituitary neuroendocrine tumors (IPitNETs) pose a significant challenge to clinical treatment due to their low rate of complete surgical resection and high recurrence rate, severely impacting patient survival and prognosis. Despite their high incidence, the molecular drivers of their pathogenesis remain poorly understood.

[0003] Recent studies have highlighted protein kinases as key regulators of tumor behavior, modulating cellular processes through complex phosphorylation networks. Cyclin-dependent kinase 5 (CDK5) and PDZ-binding kinase (PBK) have been identified as key mediators of pituitary tumor invasiveness in previous research. Importantly, building on this foundation, this study reveals a novel bidirectional phosphorylation relationship between CDK5 and PBK—this bidirectional regulation powerfully drives tumor progression.

[0004] This discovery regarding the cross-phosphorylation of CDK5 and PBK provides a highly attractive therapeutic target for pheochromocytic neuroendocrine tumors (PitNETs). However, translating these molecular mechanisms into effective treatments remains a significant challenge. Current drug therapies for PitNETs have limitations: dopamine receptor agonists, used as first-line treatment for prolactinomas, only elicit suboptimal efficacy or resistance in 10%-30% of patients. For other PitNET subtypes, existing drug treatments are mostly limited to postoperative adjuvant or salvage therapy, highlighting the critical need in clinical practice to develop novel strategies to overcome resistance and improve treatment outcomes for various subtypes.

[0005] Artificial intelligence (AI) has shown revolutionary potential in this field, significantly accelerating drug development. Based on this research framework, this invention integrates an AI-driven discovery strategy, successfully identifying an inhibitor that blocks the CDK5-PBK bidirectional phosphorylation loop. Summary of the Invention

[0006] To evaluate the therapeutic potential of simultaneously targeting CDK5 and PBK in inhibiting pituitary tumor proliferation, this invention provides an artificial intelligence-based method for screening CDK5 and PBK inhibitors in pituitary neuroendocrine tumors. The method employs an AI-based virtual screening strategy to identify dual inhibitors, constructs robust classification models that can accurately predict CDK5 / PBK inhibitory activity, and utilizes these models to discover novel dual-target compounds.

[0007] The technical solution adopted is as follows: A method for screening CDK5 and PBK inhibitors in pituitary neuroendocrine tumors based on artificial intelligence, the method comprising the following steps: Step 1: Obtain a selected inhibitor dataset of CDK5 / PBK from the ChEMBL database and divide it into training set, validation set and test set; Step 2: Several prediction models are established using machine learning algorithms and deep learning frameworks that couple molecular fingerprints and molecular graphs, and each prediction model is trained using a training set. Step 3: Optimize the hyperparameters of each prediction model using the validation set, and evaluate the prediction performance and generalization ability of each prediction model using the test set. Step 4: Select the prediction model with better robustness and prediction accuracy as the inhibitor drug screening model. Step 5: Use inhibitor drug screening models to predict molecular inhibitory activity from FDA-approved drug databases and natural compound databases; Step 6: Select molecules with prediction confidence scores greater than 0.6 for both CDK5 and PBK, sort them according to the average score, and select a variety of FDA-approved drugs and the highest-scoring natural compounds for evaluation to obtain CDK5 and PBK inhibitor drugs.

[0008] Specifically, datasets of CDK5 and PBK related compounds were collected from the ChEMBL database and processed according to the following steps: Step 1.1: Only compounds containing experimental data on CDK5 or PBK inhibition were retained; Step 1.2: Retain molecules with detailed experimental values ​​for half-inhibitory concentration, half-effective concentration EC50, inhibition constant K, and affinity constant Ka; Step 1.3: Remove duplicates; Step 1.4: The compounds in the CDK5 and PBK datasets are divided into inhibitors (≤ 1 μM) and non-inhibitors (> 1 μM).

[0009] In step 1.2, if a compound has multiple test values, the average of its reported activity values ​​is calculated as the final experimental value.

[0010] Preferably, the inhibitor dataset in step 1 is randomly divided into a training set, a validation set, and a test set, wherein the ratio of the training set, validation set, and test set is 8:1:1.

[0011] The prediction models constructed in step 2 for predicting the activity of CDK5 and PBK inhibitors include: XGBoost prediction model, SVM prediction model, GCN prediction model, GAT prediction model and AttentiverFP prediction model, and the hyperparameters of each prediction model are optimized by grid search.

[0012] In step 4, a trained GCN prediction model is used to infer potential CDK5 and PBK inhibitor drugs from the FDA-approved drug database and the natural compound small molecule database.

[0013] Preferably, the selected CDK5 and PBK inhibitors are dual-target inhibitors, including proanenierl hydrochloride, which is used to inhibit the insulin signaling pathway.

[0014] The use of an AI-based CDK5 and PBK inhibitor drug in the preparation of drugs for treating growth hormone-secreting adenomas, drug-resistant prolactin-secreting tumors, corticotropin-secreting adenomas, and thyroid-stimulating hormone-secreting adenomas.

[0015] The half-maximal inhibitory concentrations (WMCs) of the tumor-suppressing drug 5 (proanil hydrochloride) were 3.61 μM and 12.89 μM in growth hormone-secreting adenomas GH-1 and GH-2, respectively; the WMC was 11.79 μM in prolactin-resistant tumor PRL-1; 8.67 μM and 6.87 μM in corticotropin-secreting adenomas SCA-1 and SCA-2; and 12.51 μM in thyroid-stimulating hormone-secreting adenomas TSH-1.

[0016] The technical solution of the present invention has the following advantages: This invention utilizes AI-driven drug screening to identify a dual-target inhibitor, proantiazoline hydrochloride (PDPH), which exhibits potent antitumor activity against both CDK5 and PBK in vitro and in vivo. In vivo studies using a xenograft model validated the compound's efficacy in inhibiting tumor progression. By identifying the co-expression of phosphorylated CDK5 (S159) and PBK (T9) as clinically relevant biomarkers of tumor invasiveness, the development of this dual-target inhibitor specifically addresses the key limitation of compensatory kinase activation observed in traditional single-target therapies. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the specific embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the method provided by the present invention; Figure 2 This invention relates to the establishment and application of a drug screening model for CDK5 and PBK inhibitors. Wherein: (A) Schematic diagram of the model building process, (BC) Evaluation metrics of the CDK5 and PBK inhibitor prediction model, (D) Application of the trained graph convolutional network (GCN) model to infer potential CDK5 and PBK dual inhibitors from a database of FDA-approved drugs and natural compounds.

[0019] Figure 3 The survival rate of GH3 and MMQ cells after 24 hours of treatment with three different concentrations of carcinogenic compounds; Wherein: (a–b) Survival rates of GH3 cells after 24 hours of exposure to 19 different carcinogenic compounds at concentrations of 0.1 μM, 1.0 μM, and 10.0 μM. (a) GH3_D1-D9 represents replicates or batches D1–D9, while (b) GH3_D10-D19 represents D10–D19. The bar chart shows the mean survival percentage and standard deviation, with the red dashed line representing the untreated control group. (Right panel) The corresponding heatmap shows the survival rate changes at all compounds and concentrations.

[0020] (c–d) MMQ cells were also processed and analyzed. (c) MMQ_D1-D9 and (d) MMQ_D10-D19 represent two replicates. Cell survival trends were similar to those of GH3, but cytotoxic responses varied depending on the compound type and dosage. The right figure shows a heatmap representation of the data.

[0021] Figure 4 The present invention provides a dual-targeting CDK5–PBK method to reduce PitNETs cell viability and remodel insulin signaling (IRS1–AKT–mTOR).

[0022] Among them: (a–d) The dose-response survival curves of GH3 and MMQ cells after 24 hours and 48 hours of treatment with four candidate inhibitors (OncoCompound2 / 5 / 9 / 13) showed that cell activity decreased in a time- and dose-dependent manner. The mean ± standard deviation was plotted in the figure, and the half-inhibitory concentration was obtained by non-linear fitting. (e–f) shRNA knockdown of CDK5 or PBK shifted the curve of the antitumor drug 5 (proanil hydrochloride) to higher concentrations in both cell lines, consistent with its targeting effect; Figure 5 The antitumor drug 5 (proanil hydrochloride) provided by this invention inhibits pituitary tumor growth in vitro and in vivo by inducing apoptosis and reactive oxygen species generation, and shows differential sensitivity in different adenoma subtypes.

[0023] Among them: (AB) Western blot analysis showed the expression of the apoptosis marker activated caspase 3 in GH3 and MMQ cells treated with 0, 2.5, 5 and 10 μM antitumor drug 5 (proanil hydrochloride); (c) Flow cytometry analysis showed that, compared with the control group (red), the level of reactive oxygen species (ROS) was increased in GH3 cells treated with the antitumor drug 5 (proanil hydrochloride) (blue). (d) Focal formation assays performed on GH3 cells treated with different concentrations (2.5 µM, 5 µM and 10 µM) of the antitumor drug 5 (proanil hydrochloride), as well as the results of combined treatment with NAC (ROS scavenger) and ZAD (all-caspase inhibitor), showed that focal formation was reduced in the antitumor drug 5 (proanil hydrochloride) treatment group; (e) Representative images of GH3 xenograft tumors in the control group and the group treated with tumor suppressor drug 5 (proanil hydrochloride); (f) Tumor weight analysis showed that, compared with the control group, the antitumor drug 5 (proanil hydrochloride) significantly reduced tumor weight (44.9±14.0 mg vs 171.2±20.2 mg, P<0.0001). (g) Changes in mouse body weight during treatment showed no significant weight loss in the tumor suppressor drug 5 (proanil hydrochloride) treatment group. Half-maximum dose-response curves for various pituitary adenoma samples, including GH-1 (3.61 µM), GH-2 (12.89 µM), PRL-1 (11.79 µM), SCA-1 (8.67 µM), SCA-2 (6.87 µM), and TSH-1 (12.51 µM) tumors, highlighted the varying sensitivities to tumor suppressor drug 5 (proanil hydrochloride). Detailed Implementation

[0024] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] like Figure 1 As shown, this invention provides an artificial intelligence-based method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors, comprising the following steps:

S01

[0026] Datasets of CDK5 and PBK related compounds were collected from the ChEMBL database and processed as follows: (1) Compounds containing only CDK5 or PBK inhibition experimental data (experiment type = B) were retained; (2) Retain molecules with detailed experimental values ​​(half-inhibitory concentration, half-effective concentration EC50, inhibition constant K, and affinity constant Ka); if a compound has multiple test values, calculate the average of its reported activity values ​​as the final experimental value; (3) Remove duplicate items; (4) The compounds in the CDK5 and PBK datasets were divided into inhibitors (≤ 1 μM) and non-inhibitors (> 1 μM).

[0027] A total of 1330 CDK5 and PBK molecules were obtained. The resulting dataset was randomly divided into three subsets: training set (80%), validation set (10%), and test set (10%). The training set was used to build prediction models, the validation set was used to optimize the hyperparameters of each prediction model, and the test set was used to evaluate the predictive ability of the established prediction models.

[0028]

S02

[0029] This invention employs two classic machine learning (CML) algorithms (XGBoost and SVM) and three deep learning (DL) algorithms (GCN, GAT, and AttentiveFP) to construct a classification model to predict the inhibitory activity of CDK5 and PBK inhibitors.

[0030] The SVM prediction model was built using the Scikit-learn software (https: / / github.com / scikit-learn / scikit-learn, version 0.24.1).

[0031] The XGBoost model is built on the XGBoost Python package (https: / / github.com / dmlc / xgboost, version: 1.3.3).

[0032] DL-based models (GCN, GAT, and AttentiveFP) are developed using the DeepChem Python package (https: / / deepchem.io / ).

[0033] In addition, grid search is used to optimize the hyperparameters of each prediction model.

[0034] The goal of this invention is to construct robust classification prediction models that can accurately predict CDK5 / PBK inhibitory activity and to use these prediction models to discover novel dual-target compounds.

[0035] To this end, the present invention developed a total of ten prediction models, using two molecular representation paradigms—molecular fingerprints and molecular graphs—combining two traditional machine learning (CML) algorithms (Support Vector Machine [SVM] and Extreme Gradient Boosting [XGBoost]) and four state-of-the-art deep learning (DL) frameworks (Graph Convolutional Network [GCN], Graph Attention Network [GAT], Attention FP, and FP-GNN).

[0036]

S03

[0037] The specific hyperparameters of the prediction model constructed in this invention have been optimized through validation set performance; for details of the optimal configuration, please refer to (…). Figure 2 bc).

[0038] Evaluation metrics for the predictive models include Area Under the Receiver Operating Characteristic (AUC), Accuracy (ACC), Precision, Recall, and F1 Score, calculated on a validation set to quantify predictive performance and generalization ability. In the evaluation architecture, this invention preferably uses a GCN-based predictive model that demonstrates superior robustness and predictive accuracy on independent test sets. Specifically, the CDK5-GCN predictive model achieved an AUC of 0.9218, while the PBK-GCN predictive model achieved an AUC of 0.8333.

[0039] [S04] Select the prediction model with better robustness and prediction accuracy as the inhibitor drug screening model. The present invention preferably uses the GCN-CDK5 and GCN-PBK prediction models.

[0040]

S05

[0041] [S06] Molecules with prediction confidence scores greater than 0.6 for both CDK5 and PBK were screened, ranked according to the average score, and evaluated using a variety of FDA-approved drugs and the highest-scoring natural compounds to obtain CDK5 and PBK inhibitor drugs.

[0042] Inhibitory activities of molecules from FDA-approved drug databases and natural compound databases were predicted using GCN-CDK5 and GCN-PBK prediction models, respectively. Molecules with prediction confidence scores greater than 0.6 for both CDK5 and PBK were screened and ranked according to their average scores, as follows: Figure 2 As shown in d, 10 FDA-approved drugs and 9 of the highest-scoring natural compounds were selected for further evaluation, and potential CDK5 and PBK dual inhibitors were inferred.

[0043] The CDK5 and PBK inhibitors screened in this invention are dual-target inhibitors, including tumor suppressor 5 (proanil hydrochloride), which is used to inhibit the insulin signaling pathway. In the accompanying drawings of the specification, tumor suppressor 5 (proanil hydrochloride) is referred to as OncoCompound 5.

[0044] The mechanism and efficacy of the screened antitumor drug 5 (proanil hydrochloride) in inhibiting pituitary tumor cell proliferation are further explained below: Inhibiting CDK5 and PBK can weaken the proliferation of pituitary tumor cells. This invention used the CCK-8 assay to perform preliminary cytotoxicity screening of 19 anticancer compounds in MMQ and GH3 cell lines. Figure 3 As shown, all four compounds—anticancer compound 2, anticancer compound 5, anticancer compound 9, and anticancer compound 13—exhibited significant inhibitory activity in both cell models.

[0045] The half-maximal inhibitory concentrations (half-inhibitory concentrations) of these four compounds were then determined, and their efficacy in inhibiting CDK5 and PBK kinase activity in vitro was evaluated. Figure 4(ad). To further validate target specificity, the half-maximal inhibitory concentration (IC50) was reassessed in stable GH3 and MMQ cell lines by shRNA-mediated CDK5 or PBK knockdown. Consistent with targeting, the absence of CDK5 or PBK reduced the sensitivity of GH3 and MMQ cells to the antitumor drug 5 (proanil hydrochloride), pushing the IC50 to higher levels (in GH3 cells, the IC50 value surged from 6.855 μM in the control group to 23.29 μM in the CDK5 knockdown group and 19.48 μM in the PBK knockdown group, respectively; in MMQ cells, the IC50 value surged from 8.133 μM in the control group to 57.16 μM in the CDK5 knockdown group and 68.53 μM in the PBK knockdown group, respectively). Figure 4 Meanwhile, the direct inhibitory effect of the tumor suppressor drug 5 (proanil hydrochloride) on its kinase target was further confirmed, demonstrating that treatment with this compound led to a dose-dependent decrease in the phosphorylation levels of PBK and CDK5 in GH3 cells. Figure 4 I). These results indicate that CDK5 and PBK are the true pharmacological targets of the antitumor drug 5 (proanil hydrochloride) in these models.

[0046] Treatment with the antitumor drug 5 (proanil hydrochloride) significantly improved apoptosis markers (such as... Figure 5 a and Figure 5 (as shown in b) and intracellular reactive oxygen species (ROS) levels ( Figure 5 c).

[0047] Finally, to functionally verify that the antiproliferative effect is mediated by reactive oxygen species and apoptosis, a focal formation experiment was also performed. Figure 5 As shown in d, when used in combination with the reactive oxygen species scavenger NAC or the pan-cysteine ​​inhibitor ZAD, the dose-dependent inhibition of clonogenic ability by the antitumor drug 5 (proanil hydrochloride) was significantly weakened. Figure 5 The comprehensive evidence for CD provides a compelling mechanism: the antitumor drug 5 (proanil hydrochloride) targets reactive oxygen species-dependent apoptosis, ultimately inhibiting tumor cell proliferation.

[0048] To further evaluate the therapeutic potential of the tumor suppressor 5 (proanil hydrochloride) identified in in vitro experiments, in vivo validation was performed using a GH3 xenograft tumor model. Figure 5 e) Twenty days after inoculation, mice treated with the antitumor drug 5 (proanil hydrochloride) showed a significantly reduced tumor burden compared to the control group, with mean tumor weights of 44.9 ± 14.0 mg and 171.2 ± 20.2 mg, respectively (P < 0.0001). These results provide strong evidence that the antitumor drug 5 (proanil hydrochloride) has potent antitumor activity against pituitary tumors in vivo.

[0049] Concurrently, the cytotoxicity of the antitumor drug 5 (proanil hydrochloride) was investigated in newly resected human primary pituitary adenoma specimens from six patients. This cohort encompassed a variety of clinically and biologically distinct tumor subtypes, including two growth hormone (GH)-secreting adenomas, one prolactin-resistant (PRL)-producing tumor, two silent adrenocorticotropic hormone-secreting adenomas (SCAs), and one thyroid-stimulating hormone (TSH)-secreting adenoma. Dose-response analysis revealed heterogeneity in the sensitivity of these samples to the drug. Figure 5 As shown in g, the half-maximal inhibitory concentration (IC50) was 3.61 μM in GH-1 and 12.89 μM in GH-2; the IC50 for PRL-1 tumors was 11.79 μM, while the values ​​for SCA-1 and SCA-2 were 8.67 μM and 6.87 μM, respectively. The IC50 for TSH-1 adenomas was 12.51 μM.

[0050] The overall results show that the antitumor drug 5 (proanil hydrochloride) screened by artificial intelligence in this invention can not only inhibit the growth of pituitary tumors in vivo, but also induce differentiated cytotoxic responses against different adenoma subtypes, which highlights its potential application value in the precision targeted therapy strategy for pituitary tumors.

[0051] Any aspects not covered in this invention are applicable to existing technologies.

[0052] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors based on artificial intelligence, characterized in that, The method includes the following steps: Step 1: Obtain a selected inhibitor dataset of CDK5 / PBK from the ChEMBL database and divide it into training set, validation set and test set; Step 2: Several prediction models are established using machine learning algorithms and deep learning frameworks that couple molecular fingerprints and molecular graphs, and each prediction model is trained using a training set. Step 3: Optimize the hyperparameters of each prediction model using the validation set, and evaluate the prediction performance and generalization ability of each prediction model using the test set. Step 4: Select the prediction model with better robustness and prediction accuracy as the inhibitor drug screening model. Step 5: Use inhibitor drug screening models to predict molecular inhibitory activity from FDA-approved drug databases and natural compound databases; Step 6: Select molecules with prediction confidence scores greater than 0.6 for both CDK5 and PBK, sort them according to the average score, and select a variety of FDA-approved drugs and the highest-scoring natural compounds for evaluation to obtain CDK5 and PBK inhibitor drugs.

2. The method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors based on artificial intelligence according to claim 1, characterized in that, Datasets of CDK5 and PBK related compounds were collected from the ChEMBL database and processed according to the following steps: Step 1.1: Only compounds containing experimental data on CDK5 or PBK inhibition were retained; Step 1.2: Retain molecules with detailed experimental values ​​for half-inhibitory concentration, half-effective concentration EC50, inhibition constant K, and affinity constant Ka; Step 1.3: Remove duplicates; Step 1.4: The compounds in the CDK5 and PBK datasets are divided into inhibitors (≤ 1 μM) and non-inhibitors (> 1 μM).

3. The method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors based on artificial intelligence according to claim 2, characterized in that, In step 1.2, if a compound has multiple test values, the average of its reported activity values ​​is calculated as the final experimental value.

4. The method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors based on artificial intelligence according to claim 1, characterized in that, The inhibitor dataset in step 1 is randomly divided into a training set, a validation set, and a test set, with the ratio of the training set, validation set, and test set being 8:1:

1.

5. The method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors based on artificial intelligence according to claim 1, characterized in that, The prediction models constructed in step 2 for predicting the activity of CDK5 and PBK inhibitors include: XGBoost prediction model, SVM prediction model, GCN prediction model, GAT prediction model and AttentiverFP prediction model, and the hyperparameters of each prediction model are optimized by grid search.

6. The method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors based on artificial intelligence according to claim 5, characterized in that, In step 4, a trained GCN prediction model is used to infer potential CDK5 and PBK inhibitor drugs from the FDA-approved drug database and the natural compound small molecule database.

7. The method for screening CDK5 and PBK inhibitor drugs in pituitary neuroendocrine tumors based on artificial intelligence according to claim 6, characterized in that, The selected CDK5 and PBK inhibitors are dual-target inhibitors, including the antitumor drug 5 (proanil hydrochloride), which is used to inhibit the insulin signaling pathway.

8. The use of a CDK5 and PBK inhibitor drug in pituitary neuroendocrine tumors in the preparation of drugs for treating growth hormone-secreting adenomas, drug-resistant prolactin-secreting tumors, corticotropin-secreting adenomas, and thyroid-stimulating hormone-secreting adenomas.

9. The use of the CDK5 and PBK inhibitor drug in pituitary neuroendocrine tumors according to claim 8, characterized in that, The half-maximal inhibitory concentrations (WMCs) of the tumor-suppressing drug 5 (proanil hydrochloride) were 3.61 μM and 12.89 μM in growth hormone-secreting adenomas GH-1 and GH-2, respectively; the WMC was 11.79 μM in prolactin-resistant tumor PRL-1; 8.67 μM and 6.87 μM in corticotropin-secreting adenomas SCA-1 and SCA-2; and 12.51 μM in thyroid-stimulating hormone-secreting adenomas TSH-1.

Images