Key points are not available for this paper at this time.
Accurate and rapid diagnosis of human epidermal growth factor receptor-2 (HER2)-positive breast cancer, coupled with prediction of trastuzumab therapeutic efficacy, is critical for clinical decision-making to the patients with breast cancer. However, there is still no standard to be clinically used without suffering from inherent limitations. In this work, we propose a machine learning-assisted multifunctional biosensing platform utilizing enzyme-embedded hydrogen-bonded organic frameworks (HOFs). In this design, diverse HOFs@enzyme composites with distinct assembly configurations serve as sensitive array elements to interact with breast cancer-derived exosomes. Moreover, these interactions can modulate HOF-enzyme activity, generating diagnostic signal patterns that form unique exosomal molecular “fingerprint” profiles. Simultaneously, coordination with machine learning enables processing of complex sensor array-based data to amplify subtle differences of exosome between different subtypes of breast cancer, thereby enhancing the discriminatory capacity of this platform. By establishing reference fingerprints using exosomes from 96 training-set patients and validating classification accuracy against immunohistochemical in 76 test-set patients, the platform achieved 100% concordance in identifying the HER2-positive subtype, demonstrating exceptional discriminative capacity. Remarkably, the platform can also predict trastuzumab treatment response with 87.5% accuracy through clinical outcome correlation. So, by enabling precise exosome characterization from peripheral blood, this non-invasive liquid biopsy technology offers a transformative approach for precision oncology in HER2-positive breast cancer, overcoming critical limitations of current diagnostic paradigms.
Zeng et al. (Mon,) studied this question.