Issue with Crosslink Satisfaction for Ecm29–Rpt4 Complex Prediction

[eunseo05058]

Dear AlphaLink2 Development Team,

Hello, I am currently using AlphaLink2 to predict the structure of the Ecm29–Rpt4 protein complex.

According to prior crosslinking data, Lys6 of Rpt4 and Lys67 of Ecm29 form a validated inter-protein crosslink. Following the instructions provided in the GitHub documentation, I prepared both the FASTA file(Rpt4.txt) and the crosslink.csv(Rpt4_for_question.txt) file accordingly.

However, in the resulting prediction output:

• The specified crosslink was not satisfied, even with the crosslink distance cutoff set to 30 Å (based on DSSO).
• The predicted structures of the two proteins appeared spatially distant, suggesting no interaction between them.

(All other parameters were left at their default values.)

I would also like to note that I am running AlphaLink2 in a Google Colab Pro environment using an NVIDIA A100 GPU.
To assist with troubleshooting, I have uploaded both the FASTA file and the crosslink file used for this run.

Could you kindly advise whether this issue may be due to the input format, modeling parameters, or a potential limitation of the AlphaLink2 pipeline?
Any guidance on how to ensure that the crosslink constraint is correctly enforced in the prediction would be greatly appreciated.

Thank you very much for your time and support.

Best regards,
Eunseo Kim

Rpt4.txt

Rpt4_for_question.txt

v

[lhatsk]

Hi,
Which network weights did you use? Switching to v3 might help, also increasing the number of samples, although I'm not sure how well this is handled in the ColabFold. Unfortunately, one link, especially one with 30A might not be enough in the end. Running the target locally might improve results since there will probably be larger MSAs. As a side note, AlphaLink does not enforce crosslinks. The crosslinking information is used together with the template and MSA information to hopefully find a consensus.