Model Card for mikemayuare/SELFY-BPE-HIV

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 BPE (Byte Pair Encoding) tokenizer for tokenizing the input. The model is intended for sequence classification tasks and should be loaded with the AutoModelForSequenceClassification class. Both the model and tokenizer can be loaded using the from_pretrained method from the Hugging Face Transformers library.

Model Details

Model Description

This is a 🤗 transformers model fine-tuned on the HIV dataset from MoleculeNet. It classifies chemical compounds as either active or inactive in inhibiting HIV replication. The model takes SELFIES molecular representations as input and uses the BPE (Byte Pair Encoding) Tokenizer for tokenization. Both the model and the tokenizer can be loaded using the from_pretrained method from Hugging Face.

• 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/SELFYBPE

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, 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, and the tokenizer with the AutoTokenizer class from the same library. The inputs must be formatted as SELFIES strings.

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

from transformers import AutoTokenizer, AutoModelForSequenceClassification

# Load the tokenizer
tokenizer = AutoTokenizer.from_pretrained("mikemayuare/SELFY-BPE-HIV")

# Load the model
model = AutoModelForSequenceClassification.from_pretrained("mikemayuare/SELFY-BPE-HIV")