Model Card for Model ID

This model is fine-tuned on the HIV dataset from MoleculeNet and is designed to classify chemical compounds based on their ability to inhibit HIV replication. The input to the model is in the SELFIES (Self-referencing Embedded Strings) molecular representation format. The model uses the APE (Atom Pair Encoding) tokenizer for tokenizing the input, with the vocabulary stored in the same repository as the model under the file name tokenizer.json. The model is intended for sequence classification tasks and should be loaded with the AutoModelForSequenceClassification class.

Model Details

Model Description

This is a 🤗 transformers model fine-tuned on the HIV dataset from MoleculeNet. It classifies chemical compounds as active or inactive in terms of inhibiting HIV replication. The model takes SELFIES molecular representations as input and uses the APE Tokenizer for tokenization. The tokenizer’s vocabulary is stored in tokenizer.json, which is located in the same repository as the model.

Developed by: Miguelangel Leon
Funded by: This work was supported by national funds through FCT (Fundação para a Ciência e a Tecnologia), under the project - UIDB/04152/2020 (DOI:10.54499/UIDB/04152/2020) - Centro de Investigação em Gestão de Informação (MagIC)/NOVA IMS).
Model type: Sequence Classification
Language(s) (NLP): Not applicable (SELFIES molecular representation)
License: MIT
Finetuned from model: mikemayuare/SELFYAPE

Model Sources

Paper : Pending

Uses

Direct Use

This model can be used directly for binary classification of chemical compounds to predict their activity in inhibiting HIV replication. The inputs must be formatted as SELFIES strings.

Downstream Use

This model can be further fine-tuned for other chemical classification tasks, particularly those that use molecular representations in SELFIES format.

Out-of-Scope Use

This model is not designed for tasks outside of chemical compound classification or tasks unrelated to molecular data (e.g., NLP).

Bias, Risks, and Limitations

As this model is fine-tuned on the HIV dataset from MoleculeNet, it may not generalize well to compounds outside the dataset’s chemical space. Additionally, it is not suited for use in applications outside of chemical compound classification tasks.

Recommendations

Users should be cautious when applying this model to new chemical datasets that differ significantly from the HIV dataset. Thorough evaluation on the target dataset is recommended before deployment.

How to Get Started with the Model

To use the model for classification, it must be loaded with the AutoModelForSequenceClassification class from 🤗 transformers. The APE tokenizer is required to process the input data, which should be formatted as SELFIES.

You can load the APE tokenizer and the model with the following steps:

# Install the APETokenizer from the repository
# !git clone https://github.com/mikemayuare/apetokenizer
# Load the tokenizer
from src.apetokenizer.ape_tokenizer import APETokenizer

tokenizer = APETokenizer()
tokenizer.load_vocabulary("tokenizer.json")

# Load the model
from transformers import AutoModelForSequenceClassification

model = AutoModelForSequenceClassification.from_pretrained("mikemayuare/SELFY-APE-HIV")

mikemayuare
/

SELFY-APE-HIV