Search

Large-scale time-lapse microscopy experiments are useful to understand delivery and expression in RNA-based therapeutics. The resulting data has high dimensionality and high (but sparse) information content, making it challenging and costly to store and process. Early prediction of experimental outcome enables intelligent data management and decision making. We start from time-lapse data of HepG2 cells exposed to lipid-nanoparticles loaded with mRNA for expression of green fluorescent protein (GFP). We hypothesize that it is possible to predict if a cell will express GFP or not based on cell morphology at time-points prior to GFP expression. Here we present results on per-cell classification (GFP expression/no GFP expression) and regression (level of GFP expression) using three different approaches. In the first approach we use a convolutional neural network extracting per-cell features at each time point. We then utilize the same features combined with: a long-short-term memory (LSTM) network encoding temporal dynamics (approach 2); and time-series feature extraction using the python package tsfresh followed by principal component analysis and gradient boosting machines (approach 3), to reach a final classification or regression result. Application of the three approaches to a previously unanalyzed test set of cells showed good predictive performance of all three approaches but that accounting for the temporal dynamics via LSTMs or tsfresh led to significantly improved performance. The predictions made by the LSTM and tsfresh applications were not significantly different. The results highlight the benefit of accounting for temporal dynamics when studying drug delivery using high content imaging.

This report describes the final status of the service integration including numbers of active services provided by the consortium, associated partners and other third parties. The work described in this report addresses all areas and tasks of WP4 (i.e. Toxicology, Chemical Properties and Bioassay Databases, Omics Databases, Knowledge Bases and Data Mining, Ontology Services, Processing and Analysis, Predictive Toxicology, Workflows, Visualisation and Reporting). Because of their importance for service integration, we also reference to work performed in WP1 on case studies and in WP2 on e-infrastructure interoperability and deployment. The report includes examples of how the services provided by the consortium partners and associated partners are combined and integrated into workflows providing solutions to answer the specific case study questions. Further details are presented on how the services from the Implementation Challenge are strengthening the demonstration on the case studies. A detailed, up-to-date description of all services integrated in OpenRiskNet including the extensive categorization of the integration status (a process based on a predefined set of eight operations) is available on the project’s website. Here we will concentrate on statistics on the services integrated (e.g. categories, type, applicability domain and areas covered, targeted users or industries).

The Implementation Challenge was created to select external tools and services, especially in areas of risk assessment, not completely covered by the OpenRiskNet consortium, to be prioritised for their integration into the e-infrastructure. Third parties could apply for financial and technical support offered by OpenRiskNet partners. The selection was made by the scientific advisory board. The winners were asked to become an OpenRiskNet associated partner. The selection criteria included the tools which were complementary to the tools from within the consortium, the Technology Readiness Level (TRL), availability of APIs and other technical requirements. In total, 14 applications (20 tools or services) were received, while 10 applications (12 tools or services) were awarded and entered in the associated programme and the implementation process.

In this deliverable we present the availability of the OpenRiskNet reference system. The system is online with operational functionality and demonstrated operations in the OpenRiskNet case studies as well as within the Associated Partner Program. This deliverable is in the form of a Demonstrator, and apart from describing the reference site and examples of applications, it also summarises the recent developments in WP2 that has not been reported earlier.

This report describes the dissemination and training activities undertaken by the OpenRiskNet partners in the second half of the project (Jun 2018 - Nov 2019) and it is a continuation of Deliverable 3.4 that reported the dissemination and training activities up to month 18 (Dec 2016-May 2018). In this report we followed a similar structure, including details on the relevant activities developed within the project: organisation of webinars, workshops, training events and hackathons, participation at conferences and workshops, peer-reviewed publications, tutorials, public communication activities, as well as the interactions initiated with other EU infrastructures or projects to support uptake of the services and their sustainability.

Additional file 1. The file contains a step by step tutorial for running the CPVS API on a local system. It also explains the process of preparing new Docker images for new receptors. Secondly, the file contains various compounds used in the study. Thirdly, it includes property distribution of the known actives and inactives for the receptor 1BNU.

With the increasing amount of image data collected from biomedical experiments there is an urgent need for smarter and more effective analysis methods. Many scientific questions require analysis of image subregions related to some specific biology. Finding such regions of interest (ROIs) at low resolution and limiting the data subjected to final quantification at high resolution can reduce computational requirements and save time. In this paper we propose a three-step pipeline: First, bounding boxes for ROIs are located at low resolution. Next, ROIs are subjected to semantic segmentation into sub-regions at mid-resolution. We also estimate the confidence of the segmented sub-regions. Finally, quantitative measurements are extracted at high resolution. We use deep learning for the first two steps in the pipeline and conformal prediction for confidence assessment. We show that limiting final quantitative analysis to sub regions with high confidence reduces noise and increases separability of observed biological effects.

Poster presented at the 10th Symposium on Pharmaceutical Profiling in Drug Discovery and Development, Uppsala, Sweden on 2020-01-30.

Background: Life science is increasingly driven by Big Data analytics, and the MapReduce programming model has been proven successful for data-intensive analyses. However, current MapReduce frameworks offer poor support for reusing existing processing tools in bioinformatics pipelines. Furthermore, these frameworks do not have native support for application containers, which are becoming popular in scientific data processing. Results: Here we present MaRe, an open source programming library that introduces support for Docker containers in Apache Spark. Apache Spark and Docker are the MapReduce framework and container engine that have collected the largest open source community; thus, MaRe provides interoperability with the cutting-edge software ecosystem. We demonstrate MaRe on 2 data-intensive applications in life science, showing ease of use and scalability. Conclusions: MaRe enables scalable data-intensive processing in life science with Apache Spark and application containers. When compared with current best practices, which involve the use of workflow systems, MaRe has the advantage of providing data locality, ingestion from heterogeneous storage systems, and interactive processing. MaRe is generally applicable and available as open source software. eSSENCE

This report details the work involved in the federation of compute and data resources between the OpenRiskNet e-infrastructure and external resources. The reference environment has been designed to be capable of handling the majority of requirements for users’ wishes to deploy and run services. However specific situations demand solutions where either the computation, the data or both reside outside the OpenRiskNet e-infrastructure. This deliverable is related to Tasks 2.7 (Interconnecting virtual environment with external infrastructures) and Tasks 2.8 (Federation between virtual environments). Resource intensive analyses, such as those performed in toxicogenomics, can have CPU, memory or disk requirements that cannot be assumed to be available across all deployment scenarios. Human sequencing data may have restrictions on where it can be processed and the vast quantity of this data often predicates that it is more efficient to “bring the computation to the data”. In achieving Tasks 2.7 and 2.8, we can demonstrate how the virtual environment can utilise external infrastructure including commercial cloud providers and data stores.

OpenRiskNet is developing an e-infrastructure for predictive toxicology and safety assessment (of chemicals, pharmaceuticals, nanomaterials etc.) in the form of virtual research environments (VREs) and data, analysis and modelling services coming from the consortium but also from third parties, which can be deployed to these. This deliverable describes the support functions set up and how they were customised for specific user groups. Due to its specific purpose of being an infrastructure where public and commercial services are offered to the community, OpenRiskNet services different types of users and roles, which can be grouped into: End users (e.g. members of academia, industry, and regulatory agencies) who are defined as users who log into an OpenRiskNet VRE and consume one or more services or applications Developers who are involved in developing or setting up parts of the OpenRiskNet VRE, such as middleware, frameworks, data, or tools packaged as services or applications; in the latter case also referred to as tool provider System administrators, who are in charge of creating and managing the OpenShift environment for the VRE and deploying the basic services. We note that there will in many cases be an overlap between these roles, so that e.g. a Developer might also play the role of a System Administrator, but we will use these definitions in this document for clarity. For all these users, we provide user-driven website content, a virtual help desk and an issue tracker for technical matters and case study related information as well as an option for feature requests, all structured around specific user roles. The helpdesk offers frequently asked questions (FAQ), a Getting Started guide, and a way to ask questions in the form of filing support tickets, which are serviced by specific OpenRiskNet staff responsible for handling and delegating issues. Additionally, a wiki specifically designed for developers, service providers and system administrators is provided with information on the deployment of the infrastructure and the development of OpenRiskNet-compliant services A newly established services catalogue offers functionalities that integrate the services description with training and user support. The support infrastructure is fully functional and in operation. A set of Key Performance Indicators are proposed, including the statistics so far. An updated status report on the KPIs will be included in the next relevant deliverables. The direct access to different support functionalities: OpenRiskNet Helpdesk: https://openrisknet.freshdesk.com OpenRiskNet Wiki: https://github.com/OpenRiskNet/home/wiki OpenRiskNet Issue tracker: https://github.com/OpenRiskNet/home/issues OpenRiskNet Services catalogue https://openrisknet.org/e-infrastructure/services/ OpenRiskNet Library: https://openrisknet.org/library/ The website homepage (https://openrisknet.org/) offers a single entry point access to all these functionalities in a simple and user friendly manner.

Conformal Prediction is a framework that produces prediction intervals based on the output from a machine learning algorithm. In this paper we explore the case when training data is made up of multiple parts available in different sources that cannot be pooled. We here consider the regression case and propose a method where a conformal predictor is trained on each data source independently, and where the prediction intervals are then combined into a single interval. We call the approach Non-Disclosed Conformal Prediction (NDCP), and we evaluate it on a regression dataset from the UCI machine learning repository using support vector regression as the underlying machine learning algorithm, with varying number of data sources and sizes. The results show that the proposed method produces conservatively valid prediction intervals, and while we cannot retain the same efficiency as when all data is used, efficiency is improved through the proposed approach as compared to predicting using a single arbitrarily chosen source., Comment: Accepted to 8th Symposium on Conformal and Probabilistic Prediction with Applications, Golden Sands, Bulgaria, 2019

AstraZeneca (AZ) R&D use image based screening systems to study a variety of processes (including drug delivery; disease biology; and detecting potentially toxic off-target effects). Current assays however tend to focus on a single time point and do not interrogate biological systems over time. The importance of looking at cellular systems as ongoing processes is apparent and such experiments almost always lead to new biological and pathological insights. AZ are currently exploring RNA-based therapeutics. Although such therapies have shown significant promise, more research into the RNA delivery is needed before it can transform healthcare. One of the most promising current methods of delivery is through Lipid Nano-Particles (LNPs). In our project we have explored the extent to which successful LNP drug delivery and protein expression can be predicted in advance from changes in cell morphology through time (via bright field and cell stain image channels). The predictive modelling framework used for this work utilizes a combination of convolutional and recurrent neural networks.

This is the open peer reviewers comments and recommendations regarding the submitted GigaScience article and/or dataset.