Skip to main content
. Author manuscript; available in PMC: 2025 Feb 1.
Published in final edited form as: Sleep Health. 2023 Dec 11;10(1 Suppl):S41–S51. doi: 10.1016/j.sleh.2023.10.005

Figure 2. Tuning hyperparameters to balance specificity and sensibility.

Figure 2.

We identified rhythmic proteins based on two criteria: the ability to satisfy the false discovery rate (FDR) rhythm threshold (displayed in the left panel) and the amplitude requirement (shown in the middle panel) in at least one condition. The amplitude was determined using the 2-norm of its components. Each hyperparameter was investigated separately while keeping other parameters at default values. Additionally, proteins that showed a rhythmic pattern and did not have a significant difference in the pattern across the different conditions (based on the FDR comparison threshold), were categorized as circadian (shown in the right panel). We merged 1-harmonic (dot curve) and 2-harmonic (dot-dash curve) results by taking the union (full curve). The default parameter values used in our study were rhy_fdr = 0.01, amp_cutoff = 0.1, and com_fdr = 0.05.