Search

Accurate retention time (RT) prediction is important for spectral library-based analysis in data-independent acquisition mass spectrometry-based proteomics. The deep learning approach has demonstrated superior performance over traditional machine learning methods for this purpose. The transformer architecture is a recent development in deep learning that delivers state-of-the-art performance in many fields such as natural language processing, computer vision, and biology. We assess the performance of the transformer architecture for RT prediction using datasets from five deep learning models Prosit, DeepDIA, AutoRT, DeepPhospho, and AlphaPeptDeep. The experimental results on holdout datasets and independent datasets exhibit state-of-the-art performance of the transformer architecture. The software and evaluation datasets are publicly available for future development in the field.

Drug targeting promises to substantially enhance future therapies, for example through the focussing of chemotherapeutic drugs at the site of a tumor, thus reducing the exposure of healthy tissue to unwanted damage. Promising work on the steering of medication in the human body employs magnetic fields acting on nanoparticles made of paramagnetic materials. We develop a computational tool to aid in the optimization of the physical parameters of these particles and the magnetic configuration, estimating the fraction of particles reaching a given target site in a large patient-specific vascular system for different physiological states (heart rate, cardiac output, etc.). We demonstrate the excellent computational performance of our model by its application to the simulation of paramagnetic-nanoparticle-laden flows in a circle of Willis geometry obtained from an MRI scan. The results suggest a strong dependence of the particle density at the target site on the strength of the magnetic forcing and the velocity of the background fluid flow.

An experimental protocol describes a sequence of tasks executed to perform experimental research in biological and biomedical areas, e.g. genetics, immunology, neuroscience and virology. Such experimental protocols indicate, for each step, exactly how it should be executed, often including equipment, reagents, descriptions of critical steps, troubleshooting instructions, other kinds of tips, as well as any other information that researchers deem important for facilitating the reusability of the protocol. These protocols therefore have a clear systematic structure, but when published they are treated like any other scientific publication i.e. as a narrative text in HTML or PDF format. The formal structure is therefore not easily accessible and cannot be reused. This paper addresses this problem by extracting, representing and publishing steps from experimental protocols to make them Findable, Accessible, Interoperable, and Reusable (FAIR). Our work builds upon human annotations in combination with Named Entity Recognition delivering nanopublications. Our software toolkit, Nanotate, is based on a flexible web based annotation environment, namely Hypothes.is, the BioPortal NER web services and the nanopublications infrastructure. Our evaluation shows that our approach is viable and our tool user-friendly.