A pathologist-AI collaboration framework for enhancing diagnostic accuracies and efficiencies.

In pathology, the deployment of artificial intelligence (AI) in clinical settings is constrained by limitations in data collection and in model transparency and interpretability. Here we describe a digital pathology framework, nuclei.io, that incorporates active learning and human-in-the-loop real-time feedback for the rapid creation of diverse datasets and models. We validate the effectiveness of the framework via two crossover user studies that leveraged collaboration between the AI and the pathologist, including the identification of plasma cells in endometrial biopsies and the detection of colorectal cancer metastasis in lymph nodes. In both studies, nuclei.io yielded considerable diagnostic performance improvements. Collaboration between clinicians and AI will aid digital pathology by enhancing accuracies and efficiencies.

HPAanalyze: an R package that facilitates the retrieval and analysis of the Human Protein Atlas data.

BACKGROUND: The Human Protein Atlas (HPA) aims to map human proteins via multiple technologies including imaging, proteomics and transcriptomics. Access of the HPA data is mainly via web-based interface allowing views of individual proteins, which may not be optimal for data analysis of a gene set, or automatic retrieval of original images. RESULTS: HPAanalyze is an R package for retrieving and performing exploratory analysis of data from HPA. HPAanalyze provides functionality for importing data tables and xml files from HPA, exporting and visualizing data, as well as downloading all staining images of interest. The package is free, open source, and available via Bioconductor and GitHub. We provide examples of the use of HPAanalyze to investigate proteins altered in the deadly brain tumor glioblastoma. For example, we confirm Epidermal Growth Factor Receptor elevation and Phosphatase and Tensin Homolog loss and suggest the importance of the GTP Cyclohydrolase I/Tetrahydrobiopterin pathway. Additionally, we provide an interactive website for non-programmers to explore and visualize data without the use of R. CONCLUSIONS: HPAanalyze integrates into the R workflow with the tidyverse framework, and it can be used in combination with Bioconductor packages for easy analysis of HPA data.

Kinomic profiling of electromagnetic navigational bronchoscopy specimens: a new approach for personalized medicine.

PURPOSE: Researchers are currently seeking relevant lung cancer biomarkers in order to make informed decisions regarding therapeutic selection for patients in so-called "precision medicine." However, there are challenges to obtaining adequate lung cancer tissue for molecular analyses. Furthermore, current molecular testing of tumors at the genomic or transcriptomic level are very indirect measures of biological response to a drug, particularly for small molecule inhibitors that target kinases. Kinase activity profiling is therefore theorized to be more reflective of in vivo biology than many current molecular analysis techniques. As a result, this study seeks to prove the feasibility of combining a novel minimally invasive biopsy technique that expands the number of lesions amenable for biopsy with subsequent ex vivo kinase activity analysis. METHODS: Eight patients with lung lesions of varying location and size were biopsied using the novel electromagnetic navigational bronchoscopy (ENB) technique. Basal kinase activity (kinomic) profiles and ex vivo interrogation of samples in combination with tyrosine kinase inhibitors erlotinib, crizotinib, and lapatinib were performed by PamStation 12 microarray analysis. RESULTS: Kinomic profiling qualitatively identified patient specific kinase activity profiles as well as patient and drug specific changes in kinase activity profiles following exposure to inhibitor. Thus, the study has verified the feasibility of ENB as a method for obtaining tissue in adequate quantities for kinomic analysis and has demonstrated the possible use of this tissue acquisition and analysis technique as a method for future study of lung cancer biomarkers. CONCLUSIONS: We demonstrate the feasibility of using ENB-derived biopsies to perform kinase activity assessment in lung cancer patients.