10:00-11:00 Arnaud Vaganay, Director of Meta-Lab
and BITSS Catalyst
reproducible research = same results with the same protocol and data \cite{Peng_2009}
replicable research = similar results with the same protocol but new data
About the p-value "When a measure becomes a target, it ceases to be a good measure" Goodhart Law
Solutions proposed:
- Research teams need to skill up in statistics and data science. This can be achieved through training and more heterogeneous teams.
- Pre-register studies: specify and publish hypotheses beforehand (https://cos.io/prereg/).
- Better curate workflows (github etc.)
- Reflect on what could affect personal judgement
- Attempt a replication!
Interesting article: " Why Most Published Research Findings Are False " \cite{Ioannidis_2005}
Problems with replicating research:
- difficult to find funding
- difficult to get published
11:00-11:30 Martin Vetterli, EPFL President
11:30-12:30 Laurent Gatto, Computational Proteomics Unit, University of Cambridge, UK
openHayouHaveherself/himselfherself/hiselfresearchisHave you ever heard a scientist introducing herself/himself as a "closed scientist" ? No. So "when will 'open science' become simply 'science'? \cite{Watson_2015}.
What is open research?
Answer: research output = free to access (read) + free to use/re-use/mine + free to disseminate (publish) + inclusive
Open Science should be open as welcoming ! There is a variety of situations that researchers face. We cannot only think as privileged
\(\)Are you wondering why to work reproducibly? Here are 5 reasons: \cite{Markowetz_2015}.
No researcher is too junior to fix science \cite{Tregoning_2017}.
"If you want to go fast, go alone. If you want to go far, go together."
Workshop Day 1.
Stakeholders
Topics to discuss:
Group 1: Microbial and Neural Science
- Data sharing: many data are lost because they are not published. One way to counter that is to deposit data on data repositories. Online servers allow to publish data or protocols that are not published in journals. This increases the visibility and the number of citations of the research. However, this is time-consuming and some of these repositories are not free.
- Reproducibility : bad statistics, p-hacking, … --> barrier for science in general. Solution : Pre-registration forces you to think about the analysis of your data before collection. Data sharing for more transparency.
Group 2: Computational Biology
- Scooping: sharing data or other key information before submission could result in it being published by others. This is not an issue with preprints, having a previous proof with a timestamp most likely avoids this. Furthermore, in an open research group, members doing this practice could be more easily identified and weeded out. What should be done when scooped, how can open science tools help?
- solutions: a standardized way to "register" the research outcome if not the journal paper
- solutions: Post publication peer review: An online and public place where you can discuss and criticize research. (Pubpeer, Twitter, ...)
- solutions: Licensing
Group 3:
The quality of the open access journals are not recognized by t he conventional credit system. (e.g. EPFL does not finance to publish in hybrid open-access journals)
- pros for publishing in open access journals: more citations, further outreach of knowledge dissemination
- cons for publishing in open access journals : money, and not recognized in the conventional crediting systems
Solutions:
The importance of the reproducability is not enforced in research practices.