Hello, Could you please clarify how we should handle the standard deviation (SD) field in the Annotate All Missense submission format? Since most variant effect predictors are deterministic and don’t naturally provide uncertainty estimates, should we: Set SD to 0? Use a constant (e.g., 0.1)? Or is there a preferred method you'd recommend to generate per-variant SD values? Also, will the SD values affect the primary evaluation metrics like AUC in any way, or are they used only in supplementary analyses? Thanks!

Created by Kerem Çil kerem.cil
Hi @gaia.sage, Thank you very much for the clarification! This is helpful to know. Kind regards, Kerem
Hello @kerem.cil, Thanks for your question! We are still developing methods to better handle uncertainty in this context. For now, please use a constant value for the SD field across your predictions. The SD values will not affect the primary evaluation metrics (such as AUC); they may only be considered in supplementary analyses. Best regards, Gaia
Hi @kerem.cil. The @CAGI7ChallengesOrganizers would be best suited to answer your question and I've tagged them. Best of luck!

Clarification on standard deviation page is loading…