Independent reproducibility is essential to the generation of scientific knowledge. Optimizing experimental protocols to ensure reproducibility is an important aspect of scientific work. Genetic or pharmacological lifespan extensions are generally small compared to the inherent variability in mean lifespan even in isogenic populations housed under identical conditions. This variability makes reproducible detection of small but real effects experimentally challenging. In this study, we aimed to determine the reproducibility of C. elegans lifespan measurements under ideal conditions, in the absence of methodological errors or environmental or genetic background influences. To accomplish this, we generated a parametric model of C. elegans lifespan based on data collected from 5,026 wild-type N2 animals. We use this model to predict how different experimental practices, effect sizes, number of animals, and how different ‘shapes’ of survival curves affect the ability to reproduce real longevity effects. We find that the chances of reproducing real but small effects are exceedingly low and would require substantially more animals than are commonly used. Our results indicate that many lifespan studies are underpowered to detect reported changes and that, as a consequence, stochastic variation alone can account for many failures to reproduce longevity results. As a remedy, we provide power of detection tables that can be used as guidelines to plan experiments with statistical power to reliably detect real changes in lifespan and limit spurious false positive results. These considerations will improve best-practices in designing lifespan experiment to increase reproducibility.

In this dissertation I deal with the requirements and the analysis of the reproducibility. I set out methods based on provenance data to handle or eliminate the unavailable or changing descriptors in order to be able reproduce an – in other way – non-reproducible scientific workflow. In this way I intend to support the scientist’s community in designing and creating reproducible scientific workflows.

Reproducibility has become one of biology’s most pressing issues. This impasse has been fueled by the combined reliance on increasingly complex data analysis methods and the exponential growth of biological datasets. When considering the installation, deployment and maintenance of bioinformatic pipelines, an even more challenging picture emerges due to the lack of community standards. The effect of limited standards on reproducibility is amplified by the very diverse range of computational platforms and configurations on which these applications are expected to be applied (workstations, clusters, HPC, clouds, etc.). With no established standard at any level, diversity cannot be taken for granted.

The rate of progress in human neurosciences is limited by the inability to easily apply a wide range of analysis methods to the plethora of different datasets acquired in labs around the world. In this work, we introduce a framework for creating, testing, versioning and archiving portable applications for analyzing neuroimaging data organized and described in compliance with the Brain Imaging Data Structure (BIDS). The portability of these applications (BIDS Apps) is achieved by using container technologies that encapsulate all binary and other dependencies in one convenient package. BIDS Apps run on all three major operating systems with no need for complex setup and configuration and thanks to the richness of the BIDS standard they require little manual user input. Previous containerized data processing solutions were limited to single user environments and not compatible with most multi-tenant High Performance Computing systems. BIDS Apps overcome this limitation by taking advantage of the Singularity container technology. As a proof of concept, this work is accompanied by 22 ready to use BIDS Apps, packaging a diverse set of commonly used neuroimaging algorithms.

The Reproducibility Project: Cancer Biology launched in 2013 as an ambitious effort to scrutinize key findings in 50 cancer papers published in Nature, Science, Cell and other high-impact journals. It aims to determine what fraction of influential cancer biology studies are probably sound — a pressing question for the field. In 2012, researchers at the biotechnology firm Amgen in Thousand Oaks, California, announced that they had failed to replicate 47 of 53 landmark cancer papers2. That was widely reported, but Amgen has not identified the studies involved.

A large portion of scientific results is based on analysing and processing research data. In order for an eScience experiment to be reproducible, we need to able to identify precisely the data set which was used in a study. Considering evolving data sources this can be a challenge, as studies often use subsets which have been extracted from a potentially large parent data set. Exporting and storing subsets in multiple versions does not scale with large amounts of data sets. For tackling this challenge, the RDA Working Group on Data Citation has developed a framework and provides a set of recommendations, which allow identifying precise subsets of evolving data sources based on versioned data and timestamped queries. In this work, we describe how this method can be applied in small scale research data scenarios and how it can be implemented in large scale data facilities having access to sophisticated data infrastructure. We describe how the RDA approach improves the reproducibility of eScience experiments and we provide an overview of existing pilots and use cases in small and large scale settings.