EMBL-SciLifeLab Data Science workshop

EMBL and SciLifeLab recently launched data science training initiatives. In this session we will share our approaches to developing data science training  – our plans, our successes and our greatest challenges. An open discussion will explore areas for future collaboration and how we might work together to recognise those who dedicate their time, usually on a volunteer basis, to advanced scientific training.

Explore how to manage a portfolio of public computational tools and data services, with real-world examples from both emerging and established resources, and a panel discussion on the future challenges and opportunities for public data services

In this workshop:

• A keynote talk introduces key points of successfully managing a portfolio of data services
• One emerging and one established data service summarize their management approach
• A panel discussion including other data service leads from EMBL-EBI and SciLifeLab discuss the future challenges and opportunities for public data services

In this session we will focus on how to foster AI further (within EMBL-EBI and SciLifeLab) – What are the lessons learned during the last decade of incorporating learning-based methods using neural networks in biomedical research? Where lies the power, and what are the limitations? How do we handle bottle-necks such as lack of reliably annotated training data, and how do we avoid biases introduced by factors such as sample preparation and data collection? How can we include ‘the human in the loop’, and how can we share models?

Modern, data-intensive science requires complex and ideally reproducible workflows. Based on two opening presentations, one from a complex “research” workflow background, one from a repeatable “service” background, and a subsequent panel discussion, we want to explore

• What are the scientific demands and requirements for computational workflows?
• How to strike the right balance between reproducibility and flexibility?
• What is the current status and future aspirations for technical support of computational workflows?
• What are the opportunities for coordination / integration of internal /external workflows between SciLifeLab and EMBL?

The technical infrastructures and computational solutions provided by teams across EMBL and SciLifeLab are at the heart and the critical foundation of data science activities in both institutions. In this session, we aim to initiate a cross-organizational dialogue and to set the stage for further networking among interested stakeholders to explore key IT and technical challenges and learn more about potential solutions being explored on both sides.

The session will start with four short presentations to stimulate the open discussion that will follow. Our input from the presentations will cover a wide range of technical topics and challenges:

i) key aspects of using cloud services,

ii) opportunities to improve scientific data management via dedicated IT solutions,

iii) the challenges associated with shared analysis and transfer of (sensitive) data across centres, and

iv) exploring the needs for next-gen computational, data analytics, and management services for data science in the life sciences.

In the imaging session we will discuss workflows from image data generation to data analysis and finally publication. Which paths does microscopy data take at EMBL and SciLifeLab? How can data analysis and data management be integrated early on in the project discussions? The session will also give an insight to which kind of imaging projects are handled at both EMBL and SciLifeLab, since imaging data can have a broad range of characteristics and comes with different analysis and data management needs.

The Human Data session will focus on challenges working with federated and multidimensional human data, interfacing with national data , e.g. Genomics Data Infrastructure (GDI; Persson) enabling realisation of the European 1+ Million Genome project, and the experience of working at the interface of clinical research with industrial partners, including the Mannheim alliance (Stegle) and precompetitive industrial collaborations, e.g.OpenTargets (https://www.opentargets.org/). The session will have two presentations (Stegle, Persson) followed by a panel discussion. It aims to define the skills set, mindset, technical solutions to address shared problems of human research data.