Hello,
Is there a way to interpret the ORF score from TransDecoder?

As far as I understand from here, TransDecoder predictions and performance are dataset-dependent and cannot be generalized. However, I wonder whether there is a minimum score value one can expect from an ORF to be considered a "reliable" enzyme-coding sequence?

I am working on de novo transcriptome assembly data (short RNA-Seq) from plants and the reason why I am asking is that I observed a high number of ORF with score below 1 (see figure below, more details on this Biostar post). So I was wondering whether I could filter those out, or at least flag them as "unreliable"?