Background Traditional biological experiments for protein subcellular localization are costly and inefficient, while sequence-based methods fail to capture spatial dynamics of protein translocation. Existing deep learning models primarily rely on convolutions and lack global image integration, particularly in small-sample scenarios. Methods We propose ProteinFormer, a novel model integrating biological images with an enhanced pre-trained transformer architecture. It combines ResNet for local feature extraction and a modified transformer for global information fusion. To address data scarcity, we further develop GL-ProteinFormer, which incorporates residual learning, inductive bias, and a ConvFFN. Results ProteinFormer achieves state-of-the-art performance on the Cyto_2017 dataset for both single-label (91% <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12864_2025_12194_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\textrm{F}_{1}\)</EquationSource> </InlineEquation>-score) and multi-label (81% <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12864_2025_12194_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\textrm{F}_{1}\)</EquationSource> </InlineEquation>-score) tasks. GL-ProteinFormer demonstrates superior generalization on the limited-sample IHC_2021 dataset (81% <InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="12864_2025_12194_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(\textrm{F}_{1}\)</EquationSource> </InlineEquation>-score), with ConvFFN improving Accuracy by 4% while reducing computational costs. Conclusion ProteinFormer and its GL-ProteinFormer variant show superior performance over existing convolution-based methods. By fusing biological images with transformer-based global feature modeling, the proposed approach offers a robust and efficient solution for protein subcellular localization, especially in data-limited settings.

ProteinFormer: protein subcellular localization based on bioimages and modified pre-trained transformer

期刊论文
出版模式：

开放获取
发表日期：

2025年11月7日

查看全文

Xinyi An,
Yixin Li,
Huiping Liao,
Wanqiang Chen,
Guosheng Han,
Xianhua Xie,
Cuixiang Lin

摘要

Background

Traditional biological experiments for protein subcellular localization are costly and inefficient, while sequence-based methods fail to capture spatial dynamics of protein translocation. Existing deep learning models primarily rely on convolutions and lack global image integration, particularly in small-sample scenarios.

Methods

We propose ProteinFormer, a novel model integrating biological images with an enhanced pre-trained transformer architecture. It combines ResNet for local feature extraction and a modified transformer for global information fusion. To address data scarcity, we further develop GL-ProteinFormer, which incorporates residual learning, inductive bias, and a ConvFFN.

Results

ProteinFormer achieves state-of-the-art performance on the Cyto_2017 dataset for both single-label (91% \(\textrm{F}_{1}\) -score) and multi-label (81% \(\textrm{F}_{1}\) -score) tasks. GL-ProteinFormer demonstrates superior generalization on the limited-sample IHC_2021 dataset (81% \(\textrm{F}_{1}\) -score), with ConvFFN improving Accuracy by 4% while reducing computational costs.

Conclusion

ProteinFormer and its GL-ProteinFormer variant show superior performance over existing convolution-based methods. By fusing biological images with transformer-based global feature modeling, the proposed approach offers a robust and efficient solution for protein subcellular localization, especially in data-limited settings.