add folding example

cookbook/snippets/esmcfold.py

@@ -0,0 +1,67 @@
+import os
+
+from esm.sdk.forge import SequenceStructureForgeInferenceClient
+from esm.sdk.api import FoldingConfig
+from esm.utils.structure import input_builder
+
+
+def get_GLP1_sequence() -> str:
+    glp1_sequence = "MKTIIALSYIFCLVFADYKDDDDLEVLFQGPARPQGATVSLWETVQKWREYRRQCQRSLTEDPPPATDLFCNRTFDEYACWPDGEPGSFVNVSCPWYLPWASSVPQGHVYRFCTAEGLWLQKDNSSLPWRDLSECEESKRGERSSPEEQLLFLYIIYTVGYALSFSALVIASAILLGFRHLHCTRNYIHLNLFASFILRALSVFIKDAALKWMYSTAAQQHQWDGLLSYQDSLSCRLVFLLMQYCVAANYYWLLVEGVYLYTLLAFSVFSEQWIFRLYVSIGWGVPLLFVVPWGIVKYLYEDEGCWTRNSNMNYWLIIRLPILFAIGVNFLIFVRVICIVVSKLKANLMCKTDIKCRLAKSTLTLIPLLGTHEVIFAFVMDEHARGTLRFIKLFTELSFTSFQGLMVAILYCFVNNEVQLEFRKSWERWRLEHLHIQRDSSMKPLKCPTSSLSSGATAGSSMYTATCQASCSPAGLEVLFQGPHHHHHHH"
+    return glp1_sequence
+
+
+def get_semaglutide_peptide_sequence() -> str:
+    semaglutide_peptide = "HAEGTFTSDVSSYLEGQAAKEFIAWLVRGRG"  # From PDB: 7KI0
+    return semaglutide_peptide
+
+
+def get_input() -> input_builder.StructurePredictionInput:
+    """
+    Builds the input for folding semaglutide in complex with GLP-1 receptor
+    """
+    glp1_sequence = get_GLP1_sequence()
+    glp1_chain_id = "A"
+    peptide_sequence = get_semaglutide_peptide_sequence()
+    peptide_chain_id = "B"
+    peptide_aib_index = 1
+
+    glp1_receptor = input_builder.ProteinInput(
+        id=glp1_chain_id,
+        sequence=glp1_sequence,  # GLP-1 receptor
+    )
+    peptide = input_builder.ProteinInput(
+        id=peptide_chain_id,
+        sequence=peptide_sequence,
+        modifications=[
+            input_builder.Modification(position=peptide_aib_index, ccd="AIB")
+        ],
+    )
+    folding_input = input_builder.StructurePredictionInput(
+        sequences=[
+            glp1_receptor,
+            peptide,
+        ],
+    )
+    return folding_input
+
+
+# --- Main folding loop
+
+MODEL = "esmc-fold-2025-04"
+URL = "https://forge.evolutionaryscale.ai"
+API_TOKEN = os.environ.get("ESM_FORGE_API_TOKEN", "")
+
+client = SequenceStructureForgeInferenceClient(url=URL, model=MODEL, token=API_TOKEN)
+
+folding_config = FoldingConfig(
+    num_recycles=3,
+    num_sampling_steps=200,
+    seed=2,
+)
+inputs = get_input()
+
+result = client.fold_all_atom(inputs, config=folding_config)
+
+# export to mmcif for visualization in pymol
+with open("./folded.cif", "w") as f:
+    f.write(result.complex.to_mmcif())