Enhanced Protein Intrinsic Disorder Prediction Through Dual-View Multiscale Features and Multi-objective Evolutionary Algorithm

Shaokuan Wang; Pengshan Cui; Yining Qian; An-Yang Lu; Xianpeng Wang

cs.NEMar 6, 2026

Enhanced Protein Intrinsic Disorder Prediction Through Dual-View Multiscale Features and Multi-objective Evolutionary Algorithm

Authors: Shaokuan Wang, Pengshan Cui, Yining Qian, An-Yang Lu, Xianpeng Wang

Organizations: College of Information Science and Engineering, Northeastern University, Shenyang 110819, China · School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China · Key Laboratory of Data Analytics and Optimization for Smart Industry, Ministry of Education, Northeastern University, Shenyang 110819, China

Abstract

Intrinsically disordered regions of proteins play a crucial role in cell signaling and drug discovery. However, their high structural flexibility makes accurate residue-level prediction challenging. Existing methods often rely on single-view representations or rigid manual fusion strategies, which fail to effectively balance the complex interplay between local amino acid preferences and long-range sequence patterns. To address these limitations, we propose D2MOE, a Dual-View Multiscale Features and Multi-objective Evolutionary Algorithm, which consists of two stages. First, a dual-view multiscale feature extraction method is introduced. This method integrates evolutionary views with deep semantic views and employs multiscale extractors to capture structural information across diverse receptive fields. Second, a multi-objective evolutionary algorithm is designed to adaptively discover optimal fusion architectures. By co-evolving discrete feature selection and continuous fusion weights, the algorithm adaptively explores optimal cross-feature architectures to enhance predictive accuracy while maintaining model compactness. Experimental results across three benchmark datasets demonstrate that D2MOE consistently outperforms state-of-the-art methods. D2MOE combines the feature extraction capabilities of deep learning with the global search advantages of evolutionary algorithms, enabling efficient feature integration without manual design, and providing a robust computational tool for protein disorder prediction.

Figures & tables

Fig. 1: Overview of D2MOE. (A) Dual-view multiscale feature extraction. Protein sequences are mapped into semantic (ProtT5) and evolutionary (HMM) views, which are then passed through multiscale CNN and RNN extractors to generate complementary feature representations. (B) Adaptive feature fusion via (C) MOEA. The 12 candidate features are encoded and evolved using a multi-objective algorithm to simultaneously optimize feature selection, fusion operators (Add, Max, Min, Mul), and fusion weights.

Hyperparameter	Setting
Number of epochs	100
Dropout rate	0.2
Batch size	256
Maximum sequence length	700
Number of labels	2
Kernel sizes (multiscale CNNs)	9, 17

TABLE I: Hyperparameters used in D2MOE.

Panel of Figure 3 — Fig. 2: Performance comparison of D2MOE and seven representative predictors on three benchmark datasets. (a–c) Quantitative results on TS115, CASP12 and CB513, reporting MCC and F1 (bars) together with AUC and AUPR (lines). (d,e) ROC and PR curves on CASP12; AUC and AUPR are reported in the legends.

Panel of Figure 4 — Fig. 3: Validation of the dual-view design on TS115, CASP12 and CB513. D2MOE denotes the dual-view model that integrates ProtT5 embeddings with HHblits-HMM profiles, whereas T5 and HMM use the semantic view or the evolutionary view alone. Bars report MCC and F1 (left axis), and lines report AUC and AUPR (right axis).

Panel of Figure 5 — Fig. 4: Radar comparison of five fusion schemes (Add, Mul, Max, Min, and D2MOE) on (a) TS115, (b) CASP12, and (c) CB513 in terms of MCC, AUC, AUPR, and F1 (larger radius indicates better performance). D2MOE encloses all fixed operators on TS115 and CASP12, and remains dominant on CB513 except that Min is slightly higher on AUPR.

Datasets	TS115				CASP12				CB513
Methods	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC
RNNs-only	0.656	0.739	0.674	0.940	0.622	0.721	0.648	0.924	0.484	0.477	0.489	0.910
CNNs-only	0.671	0.749	0.690	0.951	0.618	0.721	0.646	0.917	0.492	0.483	0.498	0.906
DV-SSF	0.669	0.756	0.687	0.953	0.605	0.714	0.633	0.917	0.494	0.476	0.495	0.907
DV-BSF	0.666	0.754	0.682	0.951	0.606	0.721	0.634	0.919	0.488	0.466	0.490	0.916
D2MOE	0.677	0.773	0.696	0.954	0.646	0.739	0.667	0.925	0.498	0.463	0.503	0.907

TABLE II: Ablation experiments for multiscale feature strategy on TS115, CASP12, and CB513

Datasets	TS115				CASP12				CB513
Methods	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC
W=1	0.674	0.748	0.690	0.947	0.632	0.710	0.657	0.909	0.496	0.452	0.500	0.898
D2MOE	0.677	0.773	0.696	0.954	0.646	0.739	0.667	0.925	0.498	0.463	0.503	0.907

TABLE III: PERFORMANCE COMPARISON WITH DIFFERENTIAL EVOLUTION COEFFICIENTS

Datasets	TS115				CASP12				CB513
Methods	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC
DV-MsF-EA	0.675	0.773	0.694	0.953	0.623	0.735	0.654	0.917	0.499	0.483	0.505	0.905
D2MOE	0.677	0.773	0.696	0.954	0.646	0.739	0.667	0.925	0.498	0.463	0.503	0.907

TABLE IV: Evaluation of multi-objective (D2MOE) and single-objective (DV-MsF-EA) evolutionary fusion on TS115, CASP12, and CB513.

Datasets	TS115				CASP12				CB513
Methods	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC	MCC	AUPR	F1	AUC
HMM-CNN1	0.528	0.623	0.518	0.912	0.411	0.534	0.408	0.841	0.371	0.338	0.356	0.841
HMM-CNN2	0.606	0.692	0.618	0.932	0.539	0.624	0.554	0.869	0.424	0.389	0.429	0.857
HMM-CNN3	0.578	0.679	0.572	0.927	0.492	0.619	0.494	0.868	0.404	0.380	0.396	0.855
HMM-CNN4	0.614	0.693	0.634	0.935	0.585	0.669	0.605	0.871	0.434	0.405	0.442	0.869
T5-CNN1	0.663	0.763	0.679	0.950	0.624	0.736	0.647	0.923	0.486	0.469	0.487	0.918

TABLE V: Performance of the D2MOE model and twelve single features on TS115, CASP12, and CB513.

Fig. 5: The best fused architecture using MOEA.

Multi-Scale Feature Fusion

Multi-scale feature fusion combines representations at different resolutions or receptive fields to improve perception, prediction, and reconstruction.

See all 183 papers

Multi-objective Genetic Programming with Multi-view Multi-level Feature for Enhanced Protein Secondary Structure Prediction
Mar 11, 2026Yining Qian, Lijie Su, Meiling Xu +1Multi-View Feature Fusion Protein Structure Prediction
National Frontiers Science Center for Industrial Intelligence and Systems Optimization, Shenyang 110819, China · Key Laboratory of Data Analytics and Optimization for Smart Industry (Northeastern University), Ministry of Education, Shenyang 110819, China · Liaoning Engineering Laboratory of Data Analytics and Optimization for Smart Industry, Shenyang 110819, China +2
Protein Structure Prediction: From Evolutionary Constraints to Generative Modeling
Aug 17, 2026Wengan He, Yongsheng Luo, Lihong Jiang +2Protein Structure Prediction Protein Representation Learning
Zhuhai College of Science and Technology, Zhuhai, China · Zhuhai UNO Technology Co., Ltd., Zhuhai, China
Property-driven Protein Inverse Folding With Multi-Objective Preference Alignment
Mar 6, 2026Xiaoyang Hou, Junqi Liu, Chence Shi +3Protein Design Preference Alignment
BioGeometry · School of Computer Science, Peking University · Mila - Québec AI Institute +3

Mar 11, 2026J/K move · Enter open · S save

Multi-objective Genetic Programming with Multi-view Multi-level Feature for Enhanced Protein Secondary Structure Prediction

Yining Qian, Lijie Su, Meiling Xu, Xianpeng Wang

National Frontiers Science Center for Industrial Intelligence and Systems Optimization, Shenyang 110819, China · Key Laboratory of Data Analytics and Optimization for Smart Industry (Northeastern University), Ministry of Education, Shenyang 110819, China · Liaoning Engineering Laboratory of Data Analytics and Optimization for Smart Industry, Shenyang 110819, China+2

Multi-View Feature Fusion Protein Structure Prediction Genetic Programming Multi-Objective Evolutionary Optimization

Open paper AskPDF arXiv

Predicting protein secondary structure is essential for understanding protein function and advancing drug discovery. However, the intricate sequence-structure relationship poses significant challenges for accurate modeling. To address these, we propose MOGP-MMF, a multi-objective genetic programming framework that reformulates PSSP as an automated optimization task focused on feature selection and fusion. Specifically, MOGP-MMF introduces a multi-view multi-level representation strategy that integrates evolutionary, semantic, and newly introduced structural views to capture the comprehensive protein folding logic. Leveraging an enriched operator set, the framework evolves both linear and nonlinear fusion functions, effectively capturing high-order feature interactions while reducing fusion complexity. To resolve the accuracy-complexity trade-off, an improved multi-objective GP algorithm is developed, incorporating a knowledge transfer mechanism that utilizes prior evolutionary experience to guide the population toward global optima. Extensive experiments across seven benchmark datasets demonstrate that MOGP-MMF surpasses state-of-the-art methods, particularly in Q8 accuracy and structural integrity. Furthermore, MOGP-MMF generates a diverse set of non-dominated solutions, offering flexible model selection schemes for various practical application scenarios. The source code is available on GitHub: https://github.com/qian-ann/MOGP-MMF/tree/main.

Abstract

Multi-Scale Feature Fusion

Explore similar work

Multi-objective Genetic Programming with Multi-view Multi-level Feature for Enhanced Protein Secondary Structure Prediction

Protein Structure Prediction: From Evolutionary Constraints to Generative Modeling

Property-driven Protein Inverse Folding With Multi-Objective Preference Alignment