Interpretation of "confidance" in classification output

[LukeLikesDirt]

Hi Duong

I'm just wondering how to interpret the "confidence" measure in the classification output. Confidence for species classifications in my current dataset ranges from 0.73 to 1.

Do you have recommendations for accepting species and genus annotations based on confidence in the classification output?

Cheers
Luke

[vuthuyduong]

Hi Luke, Confidence measures here refer to the confidence levels of the optimal thresholds or similarity cut-offs predicted for sequence identification within the associated fungal clade at a given taxonomic level. These confidence measures can vary from low to high, as you have observed (e.g., from 0.7 to 1), depending on the current taxonomic classification of the fungal clades. If the delineation in some current taxonomic classification of a certain fungal clade is highly uncertain—for example, the delineation between genera in the family Debaryomycetaceae, where Candida appears everywhere—then the confidence measure for predicting a similarity cut-off for sequence identification (genus identification in the case of Debaryomycetaceae) will be very low.

Do you have recommendations for accepting species and genus annotations

based on confidence in the classification output? Yes, I believe this is the best we can achieve, regardless of the clustering based classification algorithms/tools used. If the confidence measure is low, but the similarity score is high (>99%), I should not be worried. Hope this answers your question. Cheers Duong

…

On Wed, 20 Nov 2024 at 10:56, Luke Florence ***@***.***> wrote: Hi Duong I'm just wondering how to interpret the "confidence" measure in the classification output. Confidence for species classifications in my current dataset ranges from 0.73 to 1. Do you have recommendations for accepting species and genus annotations based on confidence in the classification output? Cheers Luke — Reply to this email directly, view it on GitHub <#5>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AF6CZMTN6L6JFOXEUMV4TCD2BRMELAVCNFSM6AAAAABSEFE362VHI2DSMVQWIX3LMV43ASLTON2WKOZSGY3TKMRVHEZTAMQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>