SEPIQ 2026 Questions: HER2 antigen construct and antigen PDB availability
Dear SEPIQ 2026 Organizing Team,
I am a participant of SEPIQ 2026 Challenge. I have two questions for clarification:
- According to the FAQ on sepiq.uk, details about whether the HER2 antigen PDB will be provided will be released with the full guidelines.
Could you clarify:
- Will the monomer PDB file of HER2 extracellular domain be provided for Track A and Track B?
- If no PDB file is provided, are there any restrictions on using publicly available HER2 structures as templates, or do we need to predict the structure purely from sequence?
- The public training dataset README does not include information about the HER2 antigen construct used in panning experiments.
Could you share the exact construct information:
UniProt ID, residue range of expressed HER2 extracellular domain, expression system and tag information.
This information is important for building our modelling pipeline.
Thank you very much for your help!
Best regards
Jinhan Zhang
Dear Jinhan Zhang,
Thank you very much for participating in the SEPIQ2026 Challenge. Please find the answers to your questions below.
PDB files for the Her2 antigen will not be provided during the challenge period. However, the three-dimensional structures of the Her2-VHH complexes will be provided in PDB format when the ground truth is released.
Instead, we have provided the full amino acid sequence of the Her2 antigen used in this experiment below. As you pointed out, this Her2 antigen is an extracellular domain which may form dimers under experimental conditions. In light of your query, we will update the data repository description to include this information. Thank you very much.
MELAALCRWG LLLALLPPGA ASTQVCTGTD MKLRLPASPE THLDMLRHLY QGCQVVQGNL ELTYLPTNAS LSFLQDIQEV QGYVLIAHNQ VRQVPLQRLR IVRGTQLFED NYALAVLDNG DPLNNTTPVT GASPGGLREL QLRSLTEILK GGVLIQRNPQ LCYQDTILWK DIFHKNNQLA LTLIDTNRSR ACHPCSPMCK GSRCWGESSE DCQSLTRTVC AGGCARCKGP LPTDCCHEQC AAGCTGPKHS DCLACLHFNH SGICELHCPA LVTYNTDTFE SMPNPEGRYT FGASCVTACP YNYLSTDVGS CTLVCPLHNQ EVTAEDGTQR CEKCSKPCAR VCYGLGMEHL REVRAVTSAN IQEFAGCKKI FGSLAFLPES FDGDPASNTA PLQPEQLQVF ETLEEITGYL YISAWPDSLP DLSVFQNLQV IRGRILHNGA YSLTLQGLGI SWLGLRSLRE LGSGLALIHH NTHLCFVHTV PWDQLFRNPH QALLHTANRP EDECVGEGLA CHQLCARGHC WGPGPTQCVN CSQFLRGQEC VEECRVLQGL PREYVNARHC LPCHPECQPQ NGSVTCFGPE ADQCVACAHY KDPPFCVARC PSGVKPDLSY MPIWKFPDEE GACQ HHHHHH
Please note that there are no restrictions on the algorithms used to predict epitopes. It is perfectly acceptable to predict epitopes based on three-dimensional structure prediction (equivalent to Track B). The aim of this challenge is to make antibody drug discovery more efficient in the real world. As long as the answer is correct, we do not consider data leakage resulting from the use of publicly available data to be an issue. You are, of course, welcome to use the publicly available Her2 structure as a template for your structure prediction.
This challenge involves antigen-antibody complexes containing novel epitopes and paratopes. Compared to three-dimensional structural data, sequence-based data on antigen-antibody interactions obtained during the development of these antibodies can be acquired more extensively and at lower cost. Such data will continue to accumulate at an accelerating rate. The primary objective of Track A is to explore the potential applications of this substantial sequence-based antigen-antibody interaction data.
Please feel free to contact us if you have any other questions.
Best regards,
The Support Team