Predicting the biodegradability of active pharmaceutical ingredients using freely available in silico models: new application area or impossible?

The development of environmentally friendly alternatives to pharmaceuticals and active pharmaceutical ingredients (APIs) is increasingly necessary, because pharmaceutical agents can persist in the environment, cause ecotoxicological effects, and require high biodegradability due to increasing regulatory requirements. Experimental classification into biodegradable or non-biodegradable according to OECD Guideline 301 is reliable but very time- and resource-intensive, hindering efficient pharmaceutical development. In silico models offer a promising alternative by predicting environmental biodegradability computationally. However, their suitability for APIs remains uncertain. This study compares the performance of different in silico models for determining biodegradability of APIs. using a test data set of 50 different pharmaceuticals and APIs with experimentally validated biodegradability data (OECD 301D). The applicability domains (AD) of the models were examined, and the performance parameters - accuracy, sensitivity, and specificity - were calculated.Resultsshow that models predict non-biodegradable substances more reliably than biodegradable ones (higher specificity), indicating greater reliability in identifying persistent compounds. This suggests current models are better suited for early detection of potentially problematic, persistent APIs than for confidently confirming environmental safety. However, none of the models achieved sufficient accuracy for reliable predictions: all performance levels fell below the 80% threshold considered the minimum requirement for reliable in silico predictions under OECD Guideline 174 and ECHA practice. The current limitations highlight the urgent need for improved models with enhanced predictive power, particularly in accurately identifying biodegradable APIs. Future model development should focus on increasing accuracy, sensitivity, and robustness to support sustainable pharmaceutical design and regulatory decision-making.