Table 6.

Taxonomy of user-provided suggestions for improvement of GeMI

Data model Consider additional fields provided by GEO (e.g. platform) Allow a free schema, to include user-defined attributes (possibly, information about genotype or treatment) Add more possibilities to denote unknown attribute values: non specified, not applicable, none User interface Add information regarding the GSM in Panel B for user reference Reshape screen to provide a more comprehensive initial view of panels User corrections Aim to user corrections’ normalization by providing guidelines and value references (e.g. for life stage values) Provide feedback to the user when the sample to be annotated changes Training and re-training phases Return number and type of updated values after re-training Perform a stronger training of the model with gene expression-related datasets (now skewed toward ChIP-seq) Integrate the possibility of using parts of sentences suggested by users as ‘relevant for prediction’