OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Selformer

SELFormer is a Python transformer model that learns molecular representations from SELFIES strings and provides pre-trained models and scripts for generating molecular embeddings.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Selformer — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use SELFormer when a research workflow needs transformer-derived molecular embeddings from SELFIES strings, particularly as input features for computational chemistry analyses.

Who it is for

  • Cheminformatics researchers preparing learned molecular features for computational analyses.
  • Drug-discovery researchers exploring molecular representations for property-prediction workflows.
  • Materials-science researchers applying computational molecular analysis.

Documented uses

  • Convert a dataset of molecules written in SMILES into SELFIES notation.
  • Generate numerical embeddings for a SELFIES molecular dataset using a pre-trained model.
  • Use learned representations as features in downstream molecular-property research.
Before you choose

Requirements & limits

  • The embedding workflow expects SELFIES input; the repository includes a script to convert SMILES datasets.
  • A pre-trained model must be available locally for embedding generation.
  • The repository recommends Conda and specifies its dependencies in data/requirements.yml.
  • The project describes both pre-trained representations and separately fine-tuned classification and regression models.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Generate embeddings from a molecular dataset

Convert a SMILES dataset to SELFIES and use a pre-trained SELFormer model to write molecular embeddings.

Before you start

  • Conda installed, with the repository dependencies defined in data/requirements.yml.
  • A molecular dataset in SMILES format if conversion is needed; otherwise, a dataset in SELFIES notation.
  • A downloaded pre-trained SELFormer model.

Steps

  1. Create and activate the documented Conda environment using data/requirements.yml.

  2. If starting with SMILES, run generate_selfies.py with the input SMILES dataset path and output SELFIES dataset path.

  3. Run produce_embeddings.py with the SELFIES dataset, pre-trained model file, and desired embeddings output path.

Expected outcome

A file of SELFormer molecular embeddings for the input SELFIES dataset.

Guide limitations

  • The documented embedding command requires SELFIES input and a pre-trained model file.
  • The instructions describe embedding generation, not downstream model training or validation.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
HUBioDataLab/SELFormer · Official homepage title · retained in its source language
How we review sourcesSuggest a correction