Toxicological profile of Acovenoside A as an active pharmaceutical ingredient – prediction of missing key toxicological endpoints using in silico toxicology methodology

Abstract

Acovenoside A, a cardenolide glycoside from Acokanthera oppositifolia, demonstrates significant therapeutic potential in cardioprotection and oncology, particularly against non-small cell lung cancer (NSCLC). However, its toxicological profile requires thorough evaluation for safe pharmaceutical application. For this purpose a comprehensive in silico methods were applied, including ACD/Labs Percepta, STopTox, admetSAR 3.0, ADMETlab 3.0, ProTox 3.0, TEST 5.1.2, and VEGA QSAR, for prediction of a key toxicological endpoints (acute toxicity, potential health effects, skin and eye irritation, as well as endocrine disruption). These different methods and models were applied to build a comprehensive toxicological profile for Acovenoside A, synthesizing predictions to inform its potential risks and guide future research. The qualitative toxicity predictions using in silico tools (STopTox, admetSAR 3.0) shows specific structural fragments responsible for toxicity (toxicophores) and high probabilities (89.3–90 %) of acute toxicity depending on route of exposure. Quantitative acute toxicity predictions (Percepta, ProTox 3.0, Test 5.1.2, VEGA QSAR) indicated moderate to high toxicity, with LD₅₀ values ranging from 6.2 mg/kg (intravenous, mice) to 51 mg/kg (subcutaneous, mice), and oral administration LD₅₀ values of 5–49 mg/kg. The digoxigenin scaffold present in Acovenoside A was associated with increased toxicity, consistent with similar compounds exhibiting a median LD₅₀ of 9.2 mg/kg. Health effects assessments highlighted substantial risks of multiorgan toxicity, with high probabilities of adverse effects on the cardiovascular, gastrointestinal, respiratory, renal, hematologic, and hepatic systems. Prediction for eye irritation (Percepta, STopTox, admetSAR 3.0, ADMETlab 3.0, VEGA QSAR) suggested minimal risk, with probabilities ranging from 0 % to 39 %, though some results fell outside the domain of applicability. For skin irritation (Percepta, STopTox, admetSAR 3.0, ADMETlab 3.0, VEGA QSAR), moderate potential was predicted (30–37 %), but reliability varied across models, underscoring the need for experimental confirmation. Endocrine disruption (Percepta, admetSAR 3.0, VEGA QSAR) risk appears low, with minimal predicted binding affinity to estrogen receptors (LogRBA > −3) and inactivity in some models. This integrative analysis of multiple in silico tools provides valuable insights into the toxicological profile of Acovenoside A. While the compound holds therapeutic promise, its toxicological risks necessitate careful dosing and further experimental validation to ensure safety across various applications.