BioMatrix-4B-Base/README.md

---
license: apache-2.0
language:
- en
tags:
- biology
- chemistry
- molecule
- protein
- multimodal
- foundation-model
- pretrained
pipeline_tag: text-generation
base_model: Qwen/Qwen3-4B-Base
library_name: transformers
---

# BioMatrix-4B-Base

**BioMatrix** is a multimodal biological foundation model that natively integrates **1D sequences**, **3D structures**, and **natural language** for both **molecules** and **proteins** within a single decoder-only architecture.

This is the **4B-parameter Base model**, obtained via **multimodal continual pretraining** of Qwen3-4B-Base on 304.4 billion tokens spanning text, molecular and protein 1D/3D data, and cross-modal corpora. This base checkpoint is intended for further fine-tuning on downstream tasks. For an instruction-tuned model ready for inference, see [BioMatrix-4B-SFT](https://huggingface.co/QizhiPei/BioMatrix-4B-SFT).

- 📄 **Paper**: [BioMatrix: Towards a Comprehensive Biological Foundation Model Spanning the Modality Matrix of Sequences, Structures, and Language](https://github.com/QizhiPei/BioMatrix/blob/main/biomatrix_tech_report.pdf)
- 💻 **Code**: [https://github.com/QizhiPei/BioMatrix](https://github.com/QizhiPei/BioMatrix)
- 🤗 **Model & Data Collection**: [https://huggingface.co/collections/QizhiPei/biomatrix](https://huggingface.co/collections/QizhiPei/biomatrix)

## Model Overview

BioMatrix maps **all biological modalities into a shared discrete token space** via a unified tokenization scheme:

- **Molecular 1D sequences** (both SMILES and SELFIES notations)
- **Molecular 3D structures** (via MolStrucTok with branch-decoupled decoder)
- **Protein 1D sequences** (residue-level tokens)
- **Protein 3D structures** (via GCP-VQVAE backbone tokenizer)
- **Natural language** (inherited from Qwen3 tokenizer)

All modalities are consumed and produced uniformly under a **single next-token prediction objective**—without external encoders, projection adapters, or modality-specific output heads.

| Model | Molecule 1D | Molecule 3D | Protein 1D | Protein 3D | Natural Language |
|-------|:-----------:|:-----------:|:----------:|:----------:|:----------------:|
| ESM3 | ✗ | ✗ | ✓ | ✓ | ✓ |
| 3D-MoLM | ✓ | ✓ | ✗ | ✗ | ✓ |
| AlphaFold3 | ✓ | ✓ | ✓ | ✓ | ✗ |
| BioT5/BioT5+ | ✓ | ✗ | ✓ | ✗ | ✓ |
| BioMedGPT | ✓ | ✗ | ✓ | ✗ | ✓ |
| **BioMatrix** | **✓** | **✓** | **✓** | **✓** | **✓** |

## Model Details

- **Base Architecture**: Qwen3-4B-Base
- **Parameters**: 4B
- **Training Stage**: Multimodal Continual Pretraining only (not instruction-tuned)
- **Training Tokens**: 304.4B
- **Context Length**: 8,192 tokens
- **Tokenizer**: Extended Qwen3 vocabulary with:
  - 11,294 joint molecular 3D tokens (composed from SELFIES atom × MolStrucTok codes)
  - 4,096 protein 3D tokens (GCP-VQVAE codebook)
  - 26 protein 1D tokens (amino acids + non-standard/unknown)
  - SELFIES atom tokens and modality-specific control tokens

### Embedding Initialization

New vocabulary entries are initialized via a **description-based scheme**: each new token is grounded in the pretrained Qwen3 embedding space by averaging the embeddings of the subword tokens of a short natural-language description (e.g., `<A_W>` → "Tryptophan"), plus a small isotropic Gaussian perturbation to break symmetry. This provides a more stable starting point than random initialization.

## Pretraining Corpus (304.4B tokens)

| Category | Tokens | Sources |
|----------|--------|---------|
| **Text** (105.3B) | General: 25.6B | FineWeb-Edu |
| | Scientific: 79.7B | FineFineWeb (biology/chemistry/medical/health), PubMed Full Articles |
| **Molecule** (73.7B) | 1D: 36.0B | PubChem, MolTextNet |
| | 3D: 17.6B | PubChem, PCQM4Mv2, PubChemQC |
| | Other: 24.0B | (text descriptions, properties, IUPAC names) |
| **Protein** (77.4B) | 1D: 17.1B | UniRef50 |
| | 3D: 38.5B | RCSB PDB, AlphaFold DB |
| | Other: 19.5B | Swiss-Prot, TrEMBL annotations |
| | Other (additional): 2.9B | |
| **Cross-entity** (48.0B) | Interleaved Text: 17.1B | PubMed, bioRxiv, S2ORC, USPTO |
| | 3D: 11.4B | CrossDocked, PPIRef |
| | Other: 19.5B | BindingDB, STITCH, jglaser, AlphaSeq |

### Training Configuration

- **Framework**: LLaMA-Factory
- **Hardware**: 64 NVIDIA H100 GPUs
- **Global Batch Size**: 1,024
- **Maximum Sequence Length**: 8,192 tokens
- **Optimizer**: AdamW
- **Peak Learning Rate**: 2.0 × 10⁻⁴ (cosine schedule)
- **Warmup Steps**: 2,000
- **Total Steps**: ~36.4K (1 epoch over the full 304.4B-token corpus)

## Intended Use

This **Base model is not instruction-tuned**. It is suitable for:

- **Further fine-tuning** on custom biological tasks
- **Continued pretraining** on domain-specific corpora
- **Research on representation learning** across biomolecular modalities
- **Embedding extraction** for downstream classification/regression tasks

For ready-to-use instruction-following capabilities (e.g., molecule captioning, protein design, property prediction), please use the [SFT variant](https://huggingface.co/QizhiPei/BioMatrix-4B-SFT).

## Quick Start

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "QizhiPei/BioMatrix-4B-Base"
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True
)

# Example: Continue a SMILES sequence
prompt = "<|mol_smi_start|>CC(=O)"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=False))
```

## Modality Wrapping

When constructing inputs, biomolecular content must be wrapped with the corresponding control tokens:

| Modality | Wrapping Example |
|----------|------------------|
| Molecule SMILES | `<\|mol_smi_start\|>CC#CC#N<\|mol_smi_end\|>` |
| Molecule SELFIES | `<\|mol_sfi_start\|>[C][#C][C][#N]<\|mol_sfi_end\|>` |
| Molecule 3D | `<\|mol_3d_start\|>[H 3][C 0][#C 6]...<\|mol_3d_end\|>` |
| Protein 1D | `<\|prot_aa_start\|><A M><A R><A A>...<\|prot_aa_end\|>` |
| Protein 3D | `<\|prot_3d_start\|><S 4012><S 153><S 2091>...<\|prot_3d_end\|>` |

Natural language text is left unwrapped and serves as the default carrier modality.

## Limitations

- This model is **not instruction-tuned** and is unlikely to follow natural-language instructions out-of-the-box. Use the SFT variant for instruction-following.
- Molecular and protein 3D structures are tokenized in **disjoint geometric reference frames**, so the model cannot natively represent biomolecular complexes (e.g., docking poses).
- Heavy domain specialization may erode some general-purpose language capabilities of the underlying Qwen3 backbone.
- Coverage is limited to **small molecules and proteins**; nucleic acids, carbohydrates, and lipids are not currently supported.

## Citation

If you find BioMatrix useful, please cite:

```bibtex
@article{pei2026biomatrix,
  title={BioMatrix: Towards a Comprehensive Biological Foundation Model Spanning the Modality Matrix of Sequences, Structures, and Language},
  author={Pei, Qizhi and Zhou, Zhimeng and Duan, Yi and Zhao, Yiyang and He, Liang and Hsieh, Chang-Yu and He, Conghui and Yan, Rui and Wu, Lijun},
  year={2026}
}
```

## License

This model is released under the Apache 2.0 license. The base model (Qwen3-4B-Base) is subject to its own license terms.