AgeXtend

Introduction

AgeXtend is a multimodal, bioactivity-based, and mechanism-backed Explainable AI Predictor for geroprotectors. AgeXtend supports 2 distinct work packages: Predictor and Browser. The first work package, Predictor, involves a bioactivity-based classifier for predicting geroprotective compounds, coupled with an explainability module providing mechanistic insights into biological processes behind the predictions, a toxicity module to evaluate their toxicity in the biological context, and finally, a target module that suggests protein targets of the putative geroprotective compounds. The second work package, Browser, allows the user to explore through the 20 pre-screened databases specifically curated for these functionalities.

License Key

AgeXtend is free for academic institutions, however, for commercial utilization a commercial license key is required. Users (academic/commercial) may apply for a valid "License Key" here.

You can also generate your own predictions using AgeXtend’s Colab notebook

Environment Setup (done using requirement.txt)

Major dependencies

1. Signaturizer (v1.1.11)
2. RDKit (v2022.3.1)

Minor dependencies

1. os
2. scikit-learn (v1.2.1)
3. xgboost (v1.5.1)
4. pandas (v1.4.3)
5. numpy (v>=1.20.3)
6. tqdm
7. joblib (v1.1.1)
8. matplotlib (v>=3.2.2)
9. seaborn (v0.11.2)
10. importlib
11. importlib-resources (v5.7.1)

Quick setup

The file requirement.txt will be used for the environment setup and by running the following command.

$ pip install -r requirement.txt

How to use AgeXtend?

Installation using pip

$ pip install -i https://pypi.org/simple/AgeXtend

License activation (One time)

To apply for the license click here

>>> from AgeXtend import Predictor 

Activate AgeXtend license

>>> Predictor.license('license key') #Example: Predictor.license('KKKVFZ41111WF6RTQ')

Work Packages

AgeXtend supports 2 distinct work packages:

1. Predictor
2. Browser

Predictor

Prediction and Explainability Module

Predicts the anti-aging potential for the input SMILES:

>>> from AgeXtend import Predictor

Prepare a list of canonical SMILES (Open Babel generated) strings

>>> smiles = ['ClCC=C', 'C=CCOC(=O)CC(C)C'] 

Create an AgeXtend type object to featurize the query compounds (i.e. list of canonical SMILES)

>>> agex_obj = Predictor.featurize(smiles)

Use the AgeXtend object for predictions

>>> output = Predictor.predict(agex_obj)

Get the list of resulting data frames that are part of the rest of the modules

>>> output.keys()
dict_keys(['Anti_Aging_Prediction', 'Explainability_Status', 'Explainability_Probabilities'])

Note: Explainability_Probabilities will be empty by default unless selected otherwise by supplying additional arguments.

Get the result of a specific output data frame

>>> output['Explainability_Status']

Sub Explainability and Target Module

Additional arguments:

AgeXtend also supports the following modules along with the Prediction and Explainability modules, which can be chosen using the boolean option (True)

Example

>>> output = Predictor.predict(agex_obj, TS=True)
>>> output.keys()
dict_keys(['Anti_Aging_Prediction', 'Explainability_Status', 'Explainability_Probabilities', 'Explainability_response'])
>>> output = Predictor.predict(agex_obj, HC=True, BDL=True)
>>> output.keys()
dict_keys(['Anti_Aging_Prediction', 'Explainability_Toxicity_Status', 'Explainability_Toxicity_Probabilities', 'Druggability_and_Potential_Targets'])

Efficient Bulk Prediction / Custom AgeXtend Database

Predictor module also provides functionalities for efficient bulk data prediction

>>> Predictor.bulk_predict(input=smiles_list) 

Or user can also use pre-calculated AgeXtend type object as input

>>> agex_obj = Predictor.featurize(smiles_list)
>>> Predictor.bulk_predict(input=agex_obj)

Additional arguments:

Note The output folder of Bulk prediction function can be used as a Custom database for input in AgeXtend Browser module

Browser

To explore the AgeXtend pre-complied predictions of various databases:

>>> from AgeXtend import Browser

Use Open Babel to Generate Canonical SMILES format of the query compound as input

>>> Browser.search(query='OC(=O)CCCc1c[nH]c2c1cccc2', output='/path/to/output/folder/')

Unzip the AgeXtendBrowserOut.zip file to visualize/print the generated report (HTML format)

Note: the report file (AgeXtend_BrowserOut.html) must be in the same folder with the images folder

Additional arguments:

AgeXtend also supports the use of locally complied Predictor module outputs (Folder containing CSV format outputs)

Example

>>> Browser.search(path='/path/to/Database/Folder/', query='OC(=O)CCCc1c[nH]c2c1cccc2', output='/path/to/output/folder/')

Pre-complied AgeXtend Prediction Databases