Inconsistency in reported results between RNADiffFold and Ufold

[wangleiofficial]

Inconsistency in reported results between RNADiffFold and Ufold

Hello,

I noticed an inconsistency in the reported benchmark results on the bpRNA-new dataset between RNADiffFold and Ufold.

In Table S3 of your paper, the results are shown as follows:

Method	Precision	Recall	F1
RNADiffFold	0.672	0.693	0.659
Ufold	0.537	0.723	0.608

However, the results reported for Ufold seem inconsistent with those in the original Ufold paper. Could you please clarify:

1. How were the Ufold results in Table S3 obtained?
   • Were they reproduced by re-running Ufold on bpRNA-new, or directly cited?
2. If cited, could you confirm the exact source and version?
3. Could differences be due to experimental settings or parameter choices?

Clarifying this would help ensure fair comparison and reproducibility.

Thank you!

[festla]

Hello,
I noticed that the part of converting data_fcn_2 (i.e., one-hot encoding) to Unet-conditioner input is missing in the code published. This refers to the 17 channels required by UFold. To ensure the completeness and correctness of the code,

here quoted the transform, but I did not find the source code to execute this step.

I want to know how do you deal with this problem, or did I miss something?