Rich metadata for richer results
Edelweiss Connect was at Bio-IT World in Boston in April to launch our new EdelweissData™ platform and the event also proved to be a great opportunity to listen to the community. The conference program promised to “separate the hype from the reality” about what burgeoning new technologies like Blockchain, AI, machine learning, etc., are actually delivering to the toxicology and pharmacy fields. One of the messages that Daniel Bachler, Principal Software Architect, took away is the “interesting disconnect between what tech providers want to talk about - i.e., ‘We are doing so much AI with so much success!’ - and what seems to be the reality on the ground”.
Daniel was particularly struck by how AI is changing the value of metadata. “Several speakers in the keynote address expressed issues around metadata handling. It seems that, very often, scientists do ok with capturing the main data of experiments, etc., but a lot of metadata is often not properly captured – and, then, it turns out that the metadata would have been important to enable interesting AI approaches. One example that came up was a paper about how keeping mice in warmer conditions than usual (30°C instead of 20°) radically changed some experimental outcomes - yet, it is almost impossible to find out at what temperature mice were kept in historical experiments as these things were only vaguely described in text, often referencing even older experiments that didn't record these settings, either.”
Daniel is part of the Edelweiss Connect team developing the EdelweissData platform to store rich metadata. He notes that our work with EU-ToxRisk, ACEnano, and NanoCommons is already informed by the need to capture information rigorously that could have later importance to research.