Small discrepancies in the documentation of oracles

[miguelgondu]

Describe the bug

I am currently creating an interface for TDC s.t. it works well with SELFIES and other representations of small molecules. In the process I've noticed the following in the documentation of oracles:

1. It invites the user to create a python env using version 3.7, which is no longer supported by Conda.
2. In the case of Sitagliptin_MPO,Zaleplon_MPO, Isomers_C7H8N2O2, and Isomers_C9H10N2O2PF2Cl, I'm getting different values for the oracle:

To Reproduce

Using PyTDC==0.4.7, run

from tdc import Oracle

oracle = Oracle(name="Sitagliptin_MPO")

# Running the oracle in the example provided by the docs:
oracle("CC(C)(C)[C@H]1CCc2c(sc(NC(=O)COc3ccc(Cl)cc3)c2C(N)=O)C1")

Expected behavior

• For Sitagliptin_MPO I get 3.34970667598234e-12 instead of 0.00562486906491877
• For Zaleplon_MPO I get 0.0019017991803329235 instead of 7.752152611462035e-05
• For Isomers_C7H8N2O2 I get 2.1987591132394053e-34 instead of 7.077155389805107e-22
• For Isomers_C9H10N2O2PF2Cl I get 1.713908431542013e-15 instead of 3.775134544279098e-11

Environment:

• OS: MacOS Sonoma 14.4.1
• Python version: 3.10.12
• TDC version: 0.4.7
• Any other relevant information: Can be replicated in Colab easily.

[amva13]

We use different pickle objects depending on the scikit version and the issue likely lies there

TDC/tdc/oracles.py

Line 51 in 1d8df15

name += "_current"

checking how to resolve

[amva13]

I'd imagine the notebook is using an older scikit version in the <= 0.24 range

[amva13]

https://tdcommons.ai/functions/oracles/
https://github.com/mims-harvard/TDC/blob/main/tutorials/TDC_105_Oracle.ipynb
https://dataverse.harvard.edu/file.xhtml?fileId=6413420&version=82.0

TDC/tdc/utils/load.py

Line 159 in 1d8df15

def dataverse_download(url, path, name, types, id=None):