vtarasv
commented
1 year ago Sure, given the correspondence between the sequences produced by AlphaFold2 and those documented in UniProt, it becomes feasible to extract domain location data from UniProt by utilising the protein accession identifier (e.g., P00519). Subsequently, the AlphaFold2-generated structure may be precisely segmented based on the parsed domain start and end positions. Please attempt the provided example below. In this specific instance, it is anticipated that one of the resulting outputs will include the feature categorised as a domain with the description of Protein kinase, encompassing the range of 'begin': '242' to 'end': '493'. It is worth noting that alternative annotations for this feature may exist, in our paper we focused on identifying protein kinase, histidine kinase, PI3K/PI4K, PIPK, AGC-kinase, or CBS. In order to isolate the kinase domain, it is necessary to selectively truncate the structure to encompass residues indexed between 242 and 493.
import xml.etree.ElementTree as ET
import urllib.request
def get_xml(upacc):
url = f'https://www.uniprot.org/uniprot/{upacc}.xml'
try:
response = urllib.request.urlopen(url).read()
except (urllib.error.HTTPError, urllib.error.URLError) as e:
with open("log/err_up_urllib.error.HTTPError|URLError.txt", "a") as f:
print(upacc, "|", e, file=f)
return
try:
root = ET.fromstring(response)
except ET.ParseError as e:
with open("log/err_up_ET.ParseError.txt", "a") as f:
print(upacc, "|", e, file=f)
return
return root
def parse_feature(child, d, upacc):
assert child.tag == "{http://uniprot.org/uniprot}feature"
type_ = child.attrib["type"]
description = ""
if "description" in child.attrib:
description = child.attrib["description"]
position, begin, end = "", "", ""
for prop in child:
if prop.tag == "{http://uniprot.org/uniprot}location":
for prop1 in prop:
if prop1.tag == "{http://uniprot.org/uniprot}position":
if "position" in prop1.attrib:
position = prop1.attrib["position"]
elif prop1.tag == "{http://uniprot.org/uniprot}begin":
if "position" in prop1.attrib:
begin = prop1.attrib["position"]
elif prop1.tag == "{http://uniprot.org/uniprot}end":
if "position" in prop1.attrib:
end = prop1.attrib["position"]
if position and not begin and not end:
begin, end = position, position
elif not position and begin and end:
pass
else:
with open("log/err_up_parse_features_no_or_mix_loc.txt", "a") as f:
print(upacc, "|", type_, "|", description, file=f)
pass
if begin and end:
if type_ not in d:
d[type_] = list()
d[type_].append({"description": description, "begin": begin, "end": end})
def parse_upacc(root, upacc):
d_feat = dict()
for entry in root:
for child in entry:
if child.tag == "{http://uniprot.org/uniprot}feature":
parse_feature(child, d_feat, upacc)
return d_feat
upacc = 'P00519'
root = get_xml(upacc )
data = parse_upacc(root, upacc)
speakstone
I noticed that when using alphafold2 to predict the protein structure, you only extracted the kinases part, can you explain how to do it in detail? I hope to do the same type of structure prediction and analysis on other datasets. The research on the role of kinases is very interesting.