The Need of Standardised Metadata to Encode Causal Relationships: Towards Safer Data-Driven Machine Learning Biological Solutions

In this paper, we discuss the importance of considering causal relations in development machine learning solutions to prevent factors hampering robustness and generalisation capacity models, such as induced biases. This issue often arises when algorithm decision is affected by confounding factors. work, argue that integration research assumptions relationships can help identify potential confounders. Together with metadata information, it enable meta-comparison data acquisition pipelines. We call for standardised meta-information practices a crucial step proper development, validation, sharing. Such include detailing process, aiming automatic actionable metadata.