What if a familiar anticonvulsant could be redesigned into a leukemia candidate? This study starts with carbamazepine, computationally engineers more than fifty analogs, and sends one—CR80—into laboratory testing. Acute lymphoblastic leukemia, or ALL, is a significant concern in both pediatric and adult demographics.

Despite one hundred fifty-six approved cancer therapies based on small molecules, only five apply to all types of leukemia. Adherence to these treatments is low due to adverse side effects, creating an urgent need to identify more effective treatment options for ALL.

The study designed over fifty analogs of carbamazepine using ligand-based and structure-based drug design methodologies, then identified CR80 for further evaluation. CR80 was synthesized with a fifty percent yield and assessed in vitro against the U-937 cell line, where it obtained an IC50 of zero point eight to one micromolar and a selectivity index of two.

ALL is the most common pediatric cancer and the second most common acute leukemia in adults, with over six thousand five hundred cases per year in the United States alone. For twenty twenty-three, the American Cancer Society estimated over six thousand five hundred new cases and almost one thousand four hundred deaths from ALL in the United States.

Sixty percent of cases occur in children, with peaks at ages two to five and another after age fifty. Recent research aims to develop drugs offering targeted treatments with fewer side effects and improved patient adherence. Computer-Aided Drug Design, or CADD, is used to identify, design, and optimize compounds through artificial intelligence and molecular modeling.

CADD can accelerate drug repositioning by identifying potential new uses for existing molecules that have passed safety and toxicity tests and are already on the market. Carbamazepine is an anticonvulsant and neuropathic pain medication used for over twenty years, and it has shown effects on hematopoietic-cell replication since nineteen ninety-five.

The pharmacophoric core of carbamazepine showed intriguing potential for treating ALL. The research used structure-based drug design and ligand-based drug design to improve the pharmacodynamics of newly designed carbamazepine analogs.

For ligand-based drug design, QSAR models were constructed using the INQA-Artificial Neural Network, and this allowed prediction of IC50 and pharmacokinetic values for new candidates. To design molecules rationally, the study searched ChEMBL, DrugBank, and PubChem for molecules with biological activity against ALL and beta-tubulin.

The chosen molecules were then taken forward using two strategies. Binding affinities for beta-tubulin were evaluated in kilocalories per mole using the PDB six-QUS crystal structure, with protein preparation following the AutoDockTools protocol. After the co-crystallized paclitaxel ligand was removed, docking was conducted with the known active ligand vincristine, followed by validation and additional energy calculations.

The structures were modeled and their energies optimized in Avogadro using the MMFF94s force field, then thirty-five drugs and fifty-eight designed analogs were docked with six-QUS in AutoDock Vina. Calculations were conducted in triplicate, the lowest-RMSD pose was averaged for each compound, and interactions and distances were visualized using Discovery Studio Suite.

The INQA-Artificial Neural Network architecture was used to create a predictive QSAR model correlating molecular descriptors with experimental IC50 values for beta-tubulin and ALL drugs. The goal was to predict the IC50 value of fifty-eight designed analogs, beginning with molecular descriptors calculated using PaDEL-Descriptor version two point twenty.

Descriptors were evaluated through Pearson correlation, filtered by correlation values between zero point two and negative zero point two, and then compared with experimental IC50 values. Six molecular descriptors became inputs, twenty-one literature IC50 values became outputs, and the selected model had a coefficient of determination exceeding zero point seven.

The study found thirty molecules specifically targeting beta-tubulin, a promising target because leukemic cells divide more rapidly than normal cells. The thirty-five molecules were used for structure-based and ligand-based drug design, and boxplot analysis eliminated fourteen molecules because of their IC50 values and poor selectivity, leaving twenty-one.

The descriptor-screening process reduced one thousand five hundred forty descriptors to one thousand one hundred by eliminating zero-value descriptors, followed by Pearson-correlation filtering. Six descriptors were selected to train the INQA-Artificial Neural Network, and ten training sessions adjusted the number of nodes until a minimum neural-network cost of R2 equal to zero point seven was achieved.

Graph one-A showed an R-squared value of zero point seven three four and included cross-validation, demonstrating the validity of the built QSAR model. Table one screens twenty-one commercial and experimental drugs using ligand- and structure-based approaches.

For each compound, the authors report beta-tubulin affinity, octanol–water partitioning, experimental and ANN-predicted IC fifty values, plus predicted hERG-blocker, Ames-toxicity, and rat oral acute-toxicity scores. Using the six-Q-U-S beta-tubulin structure, this table connects potency-related measurements with early safety profiling for compounds relevant to acute lymphoblastic leukemia.

The designed molecules’ IC50 values were predicted using the constructed QSAR-ANN model. Their affinity for beta-tubulin was calculated in kilocalories per mole, while LogP and toxicity were determined using AdmeLab software three point zero.

The selection criteria required IC50 values below ten micromolar, molecular-target affinity more negative than negative eight kilocalories per mole, and LogP values between two and four. The toxicity profile also had to match or surpass that of current alternative drugs.

Only six designed candidates exceeded the negative-eight-kilocalorie-per-mole affinity criterion. Among entries thirty-eight, forty, fifty-three, fifty-nine, seventy-five, and eighty, three—entries fifty-three, fifty-nine, and eighty—had LogP values in the promising range of two to four.

Candidate eighty had the best toxicity profile among the three selected candidates and proceeded to the synthesis stage. Table two reports LBVS and SBVS results for CBZ analogs, using CR affinity, octanol–water log P, predicted IC fifty values from an artificial neural network, and three toxicity predictions: hERG blockers, Ames toxicity, and rat oral acute toxicity.

The compounds are arranged as CR numbers twenty-two through eighty, with examples such as compound twenty-two and compound fifty-two shown across the two table sections. This matters because it places activity-related predictions and multiple toxicity flags side by side for evaluating candidate analogs.

The biological activity of the synthesized CR80 compound was evaluated in vitro using a calibration curve prepared with CR80 dissolved in ethyl acetate. A CR80 solution prepared in PBS at one-times pH seven point thirty-five was interpolated to a concentration of ten micromolar.

Cytotoxic effects were analyzed in the human lymphoma-derived tumor cell line U-937 and the healthy fibroblast line L-929 to assess selectivity. Figure one shows CR80’s concentration-response effects on U-937 cancer cells and L-929 healthy cells after forty-eight and seventy-two hours, with doxorubicin included for comparison.

Cell viability is plotted across CR80 concentrations from zero point two to one micromolar, and a red line marks fifty percent viability; significance brackets indicate differences from the untreated control. The authors highlight that, at seventy-two hours, the lower concentrations maintained a safety profile for L-929 cells, supporting a reported selectivity index of two and further in vivo evaluation.

After seventy-two hours, a decrease in cell viability close to fifty percent was evident at all concentrations, while a safety profile was maintained at zero point two, zero point four, and zero point eight micromolar compared with the untreated control. CR80 showed better results in reducing cell viability than chemotherapeutic doxorubicin treatment.

The selectivity index was equal to two, a highly favorable result regarding selectivity against the evaluated tumor line. The in vitro values predicted by the QSAR model developed with the INQA-Artificial Neural Network aligned when assessing the U-937 cell line.

However, the impact on healthy L929 cells showed toxicity at ten micromolar, while dilutions yielded excellent selectivity-index results. The combined ligand-based and structure-based approach can lead to identifying a promising new drug for in vivo trials.

New QSAR models are being developed to predict activity against cancer cells and the selection index for new lead compounds. A new candidate was developed rationally for potentially treating Acute Lymphoblastic Leukemia through the synergy between ligand-based and structure-based drug design.

The starting point was the pharmacophoric core of carbamazepine, which underwent more than fifty modifications to obtain an analog with greater affinity for beta-tubulin: negative eight point sixty-six kilocalories per mole. The candidate showed a promising IC50 value of zero point eight to one micromolar in vitro on the U-937 cell line.

CR80 emerged from combined ligand-based and structure-based design, reached the laboratory with predicted beta-tubulin affinity, and showed micromolar activity and a selectivity index of two in the reported cell-line tests.