Skip to content
/ txt2hpo Public

Python library for extracting HPO encoded phenotypes from text

License

Notifications You must be signed in to change notification settings

GeneDx/txt2hpo

Folders and files

NameName
Last commit message
Last commit date
Oct 20, 2021
Oct 19, 2021
Feb 17, 2023
Feb 1, 2020
Nov 4, 2019
May 7, 2020
Oct 20, 2021
Oct 20, 2021
Mar 13, 2020
Mar 5, 2021

Repository files navigation

txt2hpo

txt2hpo is a Python library for extracting HPO-encoded phenotypes from text. txt2hpo recognizes differences in inflection (e.g. hypotonic vs. hypotonia), handles negation and comes with a built-in medical spellchecker.

Installation

Install using pip

pip install txt2hpo

Install from GitHub

git clone https://github.com/GeneDx/txt2hpo.git
cd txt2hpo
python setup.py install

Library usage

from txt2hpo.extract import Extractor
extract = Extractor()

result = extract.hpo("patient with developmental delay and hypotonia")

print(result.hpids)


["HP:0001263", "HP:0001290"]
    

txt2hpo will attempt to correct spelling errors by default, at the cost of slower processing. This feature can be turned off by setting the correct_spelling flag to False.

from txt2hpo.extract import Extractor
extract = Extractor(correct_spelling = False)
result = extract.hpo("patient with devlopental delay and hyptonia")

print(result.hpids)

[]
 
    

txt2hpo handles negation using negspaCy. To remove negated phenotypes set remove_negated flag to True. Both the extracted and negated HPO terms can be retrieved.

from txt2hpo.extract import Extractor
extract = Extractor(remove_negated=True)
result = extract.hpo("patient has developmental delay but no hypotonia")

print(result.hpids)

["HP:0001263"]

print(result.negated_hpids)

["HP:0001252"]
    

txt2hpo picks the longest overlapping phenotype by default. To disable this feature set remove_overlapping flag to False.

from txt2hpo.extract import Extractor
extract = Extractor(remove_overlapping=False)
result = extract.hpo("patient with polycystic kidney disease")

print(result.hpids)

["HP:0000113", "HP:0000112"]


extract = Extractor(remove_overlapping=True)
result = extract.hpo("patient with polycystic kidney disease")

print(result.hpids)

["HP:0000113"]
 
    

txt2hpo outputs a valid JSON string, which contains information about extracted HPIDs, their character span and matched string.

from txt2hpo.extract import Extractor
extract = Extractor()

result = extract.hpo("patient with developmental delay and hypotonia")

print(result.json)


'[{"hpid": ["HP:0001290"], "index": [37, 46], "matched": "hypotonia"}, 
{"hpid": ["HP:0001263"], "index": [13, 32], "matched": "developmental delay"}]'