FAIR compliant database development for human microbiome data samples.

Introduction: Sharing microbiome data among researchers fosters new innovations and reduces cost for research. Practically, this means that the (meta)data will have to be standardized, transparent and readily available for researchers. The microbiome data and associated metadata will then be described with regards to composition and origin, in order to maximize the possibilities for application in various contexts of research. Here, we propose a set of tools and protocols to develop a real-time FAIR (Findable. Accessible, Interoperable and Reusable) compliant database for the handling and storage of human microbiome and host-associated data. Methods: The conflicts arising from privacy laws with respect to metadata, possible human genome sequences in the metagenome shotgun data and FAIR implementations are discussed. Alternate pathways for achieving compliance in such conflicts are analyzed. Sample traceable and sensitive microbiome data, such as DNA sequences or geolocalized metadata are identified, and the role of the GDPR (General Data Protection Regulation) data regulations are considered. For the construction of the database, procedures have been realized to make data FAIR compliant, while preserving privacy of the participants providing the data. Results and discussion: An open-source development platform, Supabase, was used to implement the microbiome database. Researchers can deploy this real-time database to access, upload, download and interact with human microbiome data in a FAIR complaint manner. In addition, a large language model (LLM) powered by ChatGPT is developed and deployed to enable knowledge dissemination and non-expert usage of the database.

Deep learning applied to breast imaging classification and segmentation with human expert intervention.

PURPOSE: Automatic classification and segmentation of tumors in breast ultrasound images enables better diagnosis and planning treatment strategies for breast cancer patients. METHODS: We collected 953 breast ultrasound images from two open-source datasets and classified them with help of an expert radiologist according to BI-RADS criteria. The data was split into normal, benign and malignant classes. We then used machine learning to develop classification and segmentation algorithms. RESULTS: We found 3.92% of the images across the open-source datasets had erroneous classifications. Post-radiologist intervention, three algorithms were developed based on the classification categories. Classification algorithms distinguished images with healthy breast tissue from those with abnormal tissue with 96% accuracy, and distinguished benign from malignant images with 85% accuracy. Both algorithms generated robust F1 and AUROC metrics. Finally, the masses within images were segmented with an 80.31% DICE score. CONCLUSIONS: Our work illustrates the potential of deep learning algorithms to improve the accuracy of breast ultrasound assessments and to facilitate automated assessments.