Rappsilber-Laboratory / AlphaLink2

AlphaLink2: Integrating crosslinking MS data into Uni-Fold-Multimer
Creative Commons Attribution 4.0 International
42 stars 11 forks source link

Inter-subunit crosslinks #24

Open jimfeng9705 opened 3 weeks ago

jimfeng9705 commented 3 weeks ago

We have a homodimeric protein with chains A and B. How to define the sequence in the fasta file to be uploaded? Do we enter 1 sequence under >A, or two (identical) sequence with the same chain name?

Per the previous threads, there is no way to define inter-subunit XLs. Any progress in this?

As the AlphaLink 2 was trained with SDA, how trustworthy can we use the DSBU data? Any way to export the crosslink data from XiView for AlphaLink??

We appreciate your team doing this crosslink guided modeling for our community!

lhatsk commented 3 weeks ago

We have a homodimeric protein with chains A and B. How to define the sequence in the fasta file to be uploaded? Do we enter 1 sequence under >A, or two (identical) sequence with the same chain name?

It depends very much on what you want to do. Atm, the system will recognize automatically that this is a homodimer. You can just put in the two sequences, e.g., >A ... >B ... If you specify your crosslinks as going from A to A, you will automatically have inter- and intra-subunit crosslinks. If you only want to have either inter or intra-subunits crosslinks it gets a little more tricky. I will soon push an update that doesn't automatically recognize the chains and then you can handle it however you want.

Per the previous threads, there is no way to define inter-subunit XLs. Any progress in this?

Do you mean ONLY inter-subunit crosslinks for homomers? This is possible but requires manual intervention.

Just as a summary: If you are ok with the ambiguity in the crosslinking data, just put in both chains separately and use a crosslinking dictionary/ file from chain A to chain A.

As the AlphaLink 2 was trained with SDA, how trustworthy can we use the DSBU data?

How many links do you have? How large is the target? You can use the data but the performance will likely be worse. Model confidence should give you some idea about the trustworthiness. If you have the compute be sure to increase the number of recycling iterations and number of samples.

Any way to export the crosslink data from XiView for AlphaLink??

Not at the moment, but it's something we should be doing and I will talk to the XiView team.

lhatsk commented 2 weeks ago

Xiview now supports exporting crosslinks in the AlphaLink2 format. We changed now the behavior for homomers in v1.1. You will now need to specify if you want to include inter and intra-subunit crosslinks. E.g., for a homodimer >A ... >A ... we will internally use two sequences A,B and you need to specify intra-subunit crosslinks for A,A and inter-subunit crosslinks for A,B.

jimfeng9705 commented 2 weeks ago

Great! Can you share sample csv and fasta files for such homodimeric protein (to be uploaded for AlphaLink2 & ColabFold)?

On Jun 14, 2024, at 05:07, lhatsk @.***> wrote:

 [EXTERNAL]

Xiviewhttps://github.com/Rappsilber-Laboratory/build-xiview/issues/106 now supports exporting crosslinks in the AlphaLink2 format. We change now the behavior for homomers. You will now need to specify if you want to include inter and intra-subunit crosslinks. E.g., for a homodimer >A ... >A ... we will internally use two sequences A,B and you need to specify intra-subunit crosslinks for A,A and inter-subunit crosslinks for A,B.

— Reply to this email directly, view it on GitHubhttps://github.com/Rappsilber-Laboratory/AlphaLink2/issues/24#issuecomment-2167792899, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BBIHPOLLGO2HBJDBBSDR4ODZHLFIBAVCNFSM6AAAAABI7VHUOOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNRXG44TEOBZHE. You are receiving this because you authored the thread.Message ID: @.***>

lhatsk commented 1 week ago

Sorry, I don't have an example. But nothing changes in terms of usage. It's just that you need to define both inter- and intra-subunit crosslinks. You have these three options, assuming a homodimer the crosslink 1,50 is ambiguous:

Inter- and intra links: 1 A 50 B 0.1 1 A 50 A 0.1 1 B 50 B 0.1

Only intra-links: 1 A 50 A 0.1 1 B 50 B 0.1

Only inter-links: 1 A 50 B 0.1

jimfeng9705 commented 1 week ago

How important is it to sort the residues (by their numbers) in the crosslinks CSV file?

On 6/22/2024 9:26 AM, lhatsk wrote: [EXTERNAL]

Sorry, I don't have an example. But nothing changes in terms of usage. It's just that you need to define both inter- and intra-subunit crosslinks. You have these three options, assuming a homodimer the crosslink 1,50 is ambiguous:

Inter- and intra links: 1 A 50 B 0.1 1 A 50 A 0.1

Only intra-links: 1 A 50 A 0.1

Only inter-links: 1 A 50 B 0.1

— Reply to this email directly, view it on GitHubhttps://github.com/Rappsilber-Laboratory/AlphaLink2/issues/24#issuecomment-2184073813, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BBIHPONWZ6UPE6T2MKJUKITZIWJUDAVCNFSM6AAAAABI7VHUOOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOBUGA3TGOBRGM. You are receiving this because you authored the thread.Message ID: @.***>

lhatsk commented 1 week ago

Order doesn't matter