The gold standard for evaluating thyroid function is measuring thyrotropin (TSH) and free thyroxine (FT4), since clinical symptoms of thyroid dysfunction are neither sufficiently sensitive nor specific.
Yet, TSH and FT4 have wide reference ranges and can exhibit discordant changes in a number of conditions. This talk describes a multi-omics study seeking to identify new biomarkers for thyroid function. The plasma metabolome and proteome of healthy young volunteers treated with levothyroxine for eight weeks was measured at multiple time points before, during, and after treatment. Machine learning was used to identify a multi-omics signature discriminating between thyrotoxicosis and euthyroidism.
After discussing the above described time-agnostic analysis of the data, this talk will introduce the concept of time-series network enrichment for leveraging more information contained in such datasets.