DrugMetric is a novel unsupervised learning framework that quantitatively assesses drug-likeness based on chemical space distance. It combines variational autoencoders (VAEs) with Gaussian Mixture Models (GMMs) to effectively differentiate drug-like molecules from non-drug molecules. The framework leverages the learning ability of VAEs and the discriminative power of GMMs to identify significant differences in drug-likeness across various datasets. Additionally, DrugMetric incorporates ensemble learning to enhance its predictive capabilities. The model has been tested on multiple datasets and consistently outperforms traditional methods like QED in scoring and classification tasks. It excels in quantifying drug-likeness and accurately distinguishing candidate drugs from non-drugs. DrugMetric has potential applications in other biochemical fields and is available as a web server for drug-likeness scoring. The model's performance is supported by extensive experiments and comparisons with other methods, demonstrating its robustness and effectiveness in drug discovery.DrugMetric is a novel unsupervised learning framework that quantitatively assesses drug-likeness based on chemical space distance. It combines variational autoencoders (VAEs) with Gaussian Mixture Models (GMMs) to effectively differentiate drug-like molecules from non-drug molecules. The framework leverages the learning ability of VAEs and the discriminative power of GMMs to identify significant differences in drug-likeness across various datasets. Additionally, DrugMetric incorporates ensemble learning to enhance its predictive capabilities. The model has been tested on multiple datasets and consistently outperforms traditional methods like QED in scoring and classification tasks. It excels in quantifying drug-likeness and accurately distinguishing candidate drugs from non-drugs. DrugMetric has potential applications in other biochemical fields and is available as a web server for drug-likeness scoring. The model's performance is supported by extensive experiments and comparisons with other methods, demonstrating its robustness and effectiveness in drug discovery.