Spalt-like transcription factor 4 (SALL4) is an oncofetal protein that has been identified to drive cancer progression in hepatocellular carcinoma (HCC) and hematological malignancies. Furthermore, a high SALL4 expression level is correlated to poor prognosis in these cancers. However, SALL4 lacks well-structured small-molecule binding pockets, making it difficult to design targeted inhibitors. SALL4-induced expression of oxidative phosphorylation (OXPHOS) genes may serve as a therapeutically targetable vulnerability in HCC through OXPHOS inhibition. Because OXPHOS functions through a set of genes with intertumoral heterogeneous expression, identifying therapeutic sensitivity to OXPHOS inhibitors may not rely on a single clear biomarker. Here, we developed a workflow that utilized molecular beacons, nucleic-acid-based, activatable sensors with high specificity to the target mRNA, delivered by nanodiamonds, to establish an artificial intelligence (AI)-assisted platform for rapid evaluation of patient-specific drug sensitivity. Specifically, when the HCC cells were treated with the nanodiamond-medicated OXPHOS biosensor, high sensitivity and specificity of the sensor allowed for improved identification of OXPHOS expression in cells. Assisted by a trained convolutional neural network, drug sensitivity of cells toward an OXPHOS inhibitor, IACS-010759, could be accurately predicted. AI-assisted OXPHOS drug sensitivity assessment could be accomplished within 1 day, enabling rapid and efficient clinical decision support for HCC treatment. The work proposed here serves as a foundation for the patient-based subtype-specific therapeutic research platform and is well suited for precision medicine.
Keywords: artificial intelligence; biosensor; deep learning; hepatocellular carcinoma; molecular beacon; nanodiamond; oxidative phosphorylation.