{"id":67025,"date":"2024-03-20T10:21:22","date_gmt":"2024-03-20T09:21:22","guid":{"rendered":"https:\/\/www.embl.org\/news\/?p=67025"},"modified":"2024-03-26T23:50:59","modified_gmt":"2024-03-26T22:50:59","slug":"data-standards-impact","status":"publish","type":"post","link":"https:\/\/www.embl.org\/news\/people-perspectives\/data-standards-impact\/","title":{"rendered":"Sense makers: how standards are enabling data reuse"},"content":{"rendered":"\n

Coauthored with Dorota Badowska<\/a><\/em><\/p>\n\n\n\n

A bit like sorting a messy pile of clothes into a neatly arranged closet, minimal information standards transform unstructured data from journal articles into structured databases. This enables researchers to \u2018mine\u2019 across multiple datasets, reuse data, and gain new insights. <\/p>\n\n\n\n

What are minimum information standards? <\/strong><\/h2>\n\n\n\n

Minimum information standards are guidelines and formats for reporting scientific data generated by high-throughput methods, such as genome sequencing. They ensure all datasets are structured in the same way, making them easy to find, verify, and analyse by researchers worldwide. Standards also provide context for datasets \u2013 for example, when, where, and how the data were generated, or what species they describe.\u00a0<\/p>\n\n\n\n

Public molecular databases<\/a>, such as the ones managed by EMBL, ensure that data generated once can be reused again and again to ask new research questions, rather than information being \u2018hidden away\u2019 on the servers of individual laboratories.<\/p>\n\n\n\n

This is an efficient approach to capturing data generated by publicly-funded science, making them easy to access. In a way, it\u2019s similar to turning paper piles into books, and systematically cataloguing them at the public library where anyone can access them. Just like libraries play a role in knowledge sharing, public data resources and minimal information standards enable researchers to access and use data generated outside their own labs. <\/p>\n\n\n\n

What makes a good minimal information standard?<\/strong><\/h2>\n\n\n\n

\u201cYou have to strike a balance between what is possible and what is practical,\u201d explained Alvis Brazma, Senior Team Leader at EMBL-EBI<\/a>, and co-author of some of the first minimal information standards published. \u201cThe people generating the data will probably say the standard requires too much information, and the people analysing the data will say it\u2019s not enough. So they have to meet somewhere in the middle. But importantly, you need to try and understand what is needed for reanalysis now and try to predict what might be needed in the future. It\u2019s not an easy task! In my experience, it\u2019s best to start with a minimum, and keep adding to it once the community is on board.\u201d<\/p>\n\n\n\n

Minimum information standards typically have two parts. First, there is a set of reporting requirements \u2013 typically presented as a table or a checklist. Second, there is an agreed data format. Information about an experiment needs to be converted into the appropriate data format for it to be submitted to the relevant database. <\/p>\n\n\n\n

Driving the development of new methods<\/strong><\/h2>\n\n\n\n

Standardised data are key to developing new methods. Every bioinformatic research method, whether it be to predict new disordered proteins, to interpret the effect of protein modifications, or to analyse bioimaging data, critically hinges on the availability and unambiguity of the data used to train the methods.<\/p>\n\n\n\n

“Minimum information standards provide context that stitches together scientific outputs into the unknowable fabric of \u2018big data\u2019,\u201d said Cy Jeffries, Staff Scientist at EMBL Hamburg<\/a> and the curator of the Small Angle Scattering Biological Data Bank (SASBSB)<\/a>. \u201cIt means that results from different scientific disciplines can be linked together, reused, and openly shared to find new patterns that we have not thought of yet, but future AI might.”<\/p>\n\n\n\n

\u201cIn the age of AI, minimal information standards and standardised databases are more important than ever because they open up the data to machine learning and AI algorithms,\u201d explained Jo McEntyre, Deputy Director of EMBL-EBI<\/a>. \u201cTake AlphaFold, for example \u2013 Google DeepMind\u2019s AI system that can accurately predict protein structures. The development of AlphaFold simply wouldn\u2019t have been possible without the decades-worth of organised, annotated public protein structure and function data in the Protein Data Bank in Europe<\/a>, and UniProt<\/a>. As with many research methods, what you get out is only as good as the data you put in.\u201d<\/p>\n\n\n\n

Many flavours of standards<\/strong><\/h2>\n\n\n\n

EMBL scientists and colleagues have contributed to the development of many minimal information standards for different data types. The standards usually follow developments in technology and improved accessibility, which result in an increase in the volume of data produced. <\/p>\n\n\n\n

Below are a few examples of minimal information standards that are now widely being used in the scientific community:<\/p>\n\n\n\n