Leveraging Image-Derived Phenotypic Measurements for Drug-Target Interaction Predictions.

In recent years, protein kinases have become some of the most significant drug targets in cancer patients. Kinases are known to regulate the activity of many human proteins, and consequently their inhibition has been used to control cancer proliferation. A significant challenge in drug discovery is the rapid and efficient identification of new small molecules. In this study, we propose a novel in silico drug discovery approach to identify kinase targets that impinge on nuclear receptor signaling with data generated using high-content analysis (HCA). A high-throughput imaging dataset was generated from an siRNA human kinome screen on engineered cells that allow direct visualization of effects on estrogen receptor-α or a chimeric progesterone receptor B binding to specific DNA. Two types of kinase descriptors are extracted from these imaging data: first, a population-median-based descriptor and second a bag-of-words (BoW) descriptor that can capture heterogeneity information in the imaging data. Using these descriptors, we provide prediction results of drug-kinase-target interactions based on single-task learning, multi-task learning, and collaborative filtering methods. The best performing model in target-based drug discovery gives an area under the receiver operating characteristic curve (AUC) of 0.86, whereas the best model in ligand-based discovery gives an AUC of 0.79. These promising results suggest that imaging-based information can be used as an additional source of information to existing virtual screening methods, thereby making the drug discovery process more time and cost efficient.

A Visually Interpretable, Dictionary-Based Approach to Imaging-Genomic Modeling, With Low-Grade Glioma as a Case Study.

Radiomics is a rapidly growing field in which sophisticated imaging features are extracted from radiology images to predict clinical outcomes/responses, genetic alterations, and other outcomes relevant to a patient's prognosis or response to therapy. This approach can effectively capture intratumor phenotypic heterogeneity by interrogating the "larger" image field, which is not possible with traditional biopsy procedures that interrogate specific subregions alone. Most models in radiomics derive numerous imaging features (eg, texture, shape, size) from a radiology data set and then learn complex nonlinear hypotheses to solve a given prediction task. This presents the challenge of visual interpretability of radiomic features necessary for effective adoption of radiomic models into the clinical decision-making process. To this end, we employed a dictionary learning approach to derive visually interpretable imaging features relevant to genetic alterations in low-grade gliomas. This model can identify regions of a medical image that potentially influence the prediction process. Using a publicly available data set of magnetic resonance imaging images from patients diagnosed with low-grade gliomas, we demonstrated that the dictionary-based model performs well in predicting 2 biomarkers of interest (1p/19q codeletion and IDH1 mutation). Furthermore, the visual regions (atoms) associated with these dictionaries show association with key molecular pathways implicated in gliomagenesis. Our results show that dictionary learning is a promising approach to obtain insights into the diagnostic process and to potentially aid radiologists in selecting physiologically relevant biopsy locations.