Prediction

@article{scriven1959explanation,
    author = "Scriven, Michael",
    title = "Explanation and Prediction in Evolutionary Theory",
    year = "1959",
    journal = "Science",
    url = "https://doi.org/10.1126/science.130.3374.477",
    doi = "10.1126/science.130.3374.477",
    number = "3374",
    pages = "477-482",
    volume = "130"
}

@misc{scriven1959explanation1,
    author = "Scriven, M",
    title = "Explanation and prediction in evolutionary theory",
    year = "1959",
    howpublished = "Science, v. 130, p. 477-482",
    note = "talkorigins\_source = {true}; raw\_reference = {Scriven, M., 1959, Explanation and prediction in evolutionary theory: Science, v. 130, p. 477-482.}"
}

@misc{s20bde9d203c6035b107fe46f9c58e7530f87af9f7,
    author = "Elith, J. and Leathwick, J.",
    title = "Species Distribution Models: Ecological Explanation and Prediction Across Space and Time",
    year = "2010",
    url = "https://www.semanticscholar.org/paper/0bde9d203c6035b107fe46f9c58e7530f87af9f7",
    is_oa = "true",
    semanticscholar_citation_count = "5842",
    semanticscholar_id = "0bde9d203c6035b107fe46f9c58e7530f87af9f7"
}

@article{doi101016jevolhumbehav201106001,
    author = "Schaafsma, S. M. and Geuze, R. and Riedstra, B. and Schiefenhövel, W. and Bouma, A. and Groothuis, T.",
    title = "Handedness in a nonindustrial society challenges the fighting hypothesis as an evolutionary explanation for left-handedness",
    year = "2012",
    journal = "Evolution and Human Behavior",
    url = "https://www.semanticscholar.org/paper/a4f7018163ca03521c7fc3321020142b5f4a4abb",
    doi = "10.1016/J.EVOLHUMBEHAV.2011.06.001",
    is_oa = "true",
    number = "2",
    pages = "94-99",
    semanticscholar_citation_count = "25",
    semanticscholar_id = "a4f7018163ca03521c7fc3321020142b5f4a4abb",
    volume = "33"
}

@article{doi101007s1053901494440,
    author = "Tennant, N.",
    title = "The logical structure of evolutionary explanation and prediction: Darwinism’s fundamental schema",
    year = "2014",
    journal = "Biology \& Philosophy",
    url = "https://www.semanticscholar.org/paper/c9e7dac94c63a0b00175304f48676cfc11ae249c",
    doi = "10.1007/S10539-014-9444-0",
    is_oa = "true",
    number = "5",
    pages = "611-655",
    semanticscholar_citation_count = "3",
    semanticscholar_id = "c9e7dac94c63a0b00175304f48676cfc11ae249c",
    volume = "29"
}

@article{tennant2014the,
    author = "Tennant, Neil",
    title = "The logical structure of evolutionary explanation and prediction: Darwinism’s fundamental schema",
    year = "2014",
    journal = "Biology \& Philosophy",
    url = "https://doi.org/10.1007/s10539-014-9444-0",
    doi = "10.1007/s10539-014-9444-0",
    number = "5",
    pages = "611-655",
    volume = "29"
}

Emotional contagion – the transfer of emotions between people – is thought to occur automatically. We test the prediction, based on evolutionary psychology, that negative, threat‐related emotions transfer more automatically than positive emotions. We introduce a new paradigm for investigating emotional contagion where participants are exposed to videos of faces that morph from neutral to angry or happy expressions. Participants watched these videos under high or low cognitive load. Participants reported more happiness in the happy condition than the anger condition and more anger in the anger condition than the happy condition, supporting our new paradigm. Participants in the happy condition were significantly happier under low compared with high load. Participants were equally angry in high and low load conditions.

@article{doi101111bjso12134,
    author = "Kelly, J. and Iannone, N. and McCarty, Megan K.",
    title = "Emotional contagion of anger is automatic: An evolutionary explanation.",
    year = "2016",
    journal = "The British journal of social psychology",
    abstract = "Emotional contagion – the transfer of emotions between people – is thought to occur automatically. We test the prediction, based on evolutionary psychology, that negative, threat‐related emotions transfer more automatically than positive emotions. We introduce a new paradigm for investigating emotional contagion where participants are exposed to videos of faces that morph from neutral to angry or happy expressions. Participants watched these videos under high or low cognitive load. Participants reported more happiness in the happy condition than the anger condition and more anger in the anger condition than the happy condition, supporting our new paradigm. Participants in the happy condition were significantly happier under low compared with high load. Participants were equally angry in high and low load conditions.",
    url = "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1111/bjso.12134",
    doi = "10.1111/bjso.12134",
    is_oa = "true",
    number = "1",
    pages = "182-191",
    semanticscholar_citation_count = "66",
    semanticscholar_id = "9320433dd9849ae0a1d1ca870fa86d773142e17d",
    volume = "55"
}

Based on evolutionary theory, we predicted that cues of resource scarcity in the environment (e.g., news of droughts or food shortages) lead people to reduce their effort and performance in physically demanding work. We tested this prediction in a 2-wave field survey among employees and replicated it experimentally in the lab. In Study 1, employees who perceived resources in the environment to be scarce reported exerting less effort when their jobs involved much (but not little) physical work. In Study 2, participants who read that resources in the environment were scarce performed worse on a task demanding more (carrying books) but not less (transcribing book titles) physical work. This result was found even though better performance increased participants’ chances of additional remuneration, and even though scarcity cues did not affect individuals’ actual ability to meet their energy needs. We discuss implications for managing effort and performance, and the potential of evolutionary psychology to explain core organizational phenomena.

@article{doi101037apl0000257,
    author = "Pitesa, Marko and Thau, Stefan",
    title = "Resource Scarcity, Effort, and Performance in Physically Demanding Jobs: An Evolutionary Explanation",
    year = "2017",
    journal = "Journal of Applied Psychology",
    abstract = "Based on evolutionary theory, we predicted that cues of resource scarcity in the environment (e.g., news of droughts or food shortages) lead people to reduce their effort and performance in physically demanding work. We tested this prediction in a 2-wave field survey among employees and replicated it experimentally in the lab. In Study 1, employees who perceived resources in the environment to be scarce reported exerting less effort when their jobs involved much (but not little) physical work. In Study 2, participants who read that resources in the environment were scarce performed worse on a task demanding more (carrying books) but not less (transcribing book titles) physical work. This result was found even though better performance increased participants’ chances of additional remuneration, and even though scarcity cues did not affect individuals’ actual ability to meet their energy needs. We discuss implications for managing effort and performance, and the potential of evolutionary psychology to explain core organizational phenomena.",
    url = "https://ink.library.smu.edu.sg/lkcsb\_research/5363",
    doi = "10.1037/apl0000257",
    is_oa = "true",
    number = "3",
    pages = "237-248",
    semanticscholar_citation_count = "25",
    semanticscholar_id = "626bcdfe2b75748484dbc3297430f02e71f551b1",
    volume = "103"
}

The frog-in-the-pan (FIP) phenomenon suggests that investors are more sensitive to abrupt price changes than gradual price changes in the stock market. Based on the cognitive-evolutionary model of surprise and the reinforcement learning model, this paper provides a new explanation for the FIP phenomenon in that this phenomenon could be explained by the elicitation of surprise emotion. We predict that when a change substantially and abruptly occurs, the significant prediction error triggers participants’ surprise, which makes participants more sensitive to the change. To ascertain these hypotheses, we recruited 109 participants and compared participants’ learning rates and surprise responses under different contexts. We observed that participants’ learning rate soared when the prediction error was large enough to trigger surprise emotion under abruptly changed conditions and confirmed that the FIP phenomenon could be explained by the elicitation of surprise emotion. In a word, this research demonstrates the significant role of surprise emotion in the decision-making process.

@article{doi103390bs13010007,
    author = "Liang, Dapeng and Liu, Mengting and Fu, Yang and Sun, Jiayin and Wang, Hongyan",
    title = "A New Explanation for the Frog-in-the-Pan Phenomenon Based on the Cognitive-Evolutionary Model of Surprise",
    year = "2022",
    journal = "Behavioral Sciences",
    abstract = "The frog-in-the-pan (FIP) phenomenon suggests that investors are more sensitive to abrupt price changes than gradual price changes in the stock market. Based on the cognitive-evolutionary model of surprise and the reinforcement learning model, this paper provides a new explanation for the FIP phenomenon in that this phenomenon could be explained by the elicitation of surprise emotion. We predict that when a change substantially and abruptly occurs, the significant prediction error triggers participants’ surprise, which makes participants more sensitive to the change. To ascertain these hypotheses, we recruited 109 participants and compared participants’ learning rates and surprise responses under different contexts. We observed that participants’ learning rate soared when the prediction error was large enough to trigger surprise emotion under abruptly changed conditions and confirmed that the FIP phenomenon could be explained by the elicitation of surprise emotion. In a word, this research demonstrates the significant role of surprise emotion in the decision-making process.",
    url = "https://www.mdpi.com/2076-328X/13/1/7/pdf?version=1672206072",
    doi = "10.3390/bs13010007",
    is_oa = "true",
    number = "1",
    pages = "7",
    semanticscholar_citation_count = "2",
    semanticscholar_id = "8cfc439d4a3988f97e89c443c5ecc5b68bae96cf",
    volume = "13"
}

Understanding the pathogenicity of missense mutation (MM) is essential for shed light on genetic diseases, gene functions, and individual variations. In this study, we propose a novel computational approach, called MMPatho, for enhancing missense mutation pathogenic prediction. First, we established a large-scale nonredundant MM benchmark data set based on the entire Ensembl database, complemented by a focused blind test set specifically for pathogenic GOF/LOF MM. Based on this data set, for each mutation, we utilized Ensembl VEP v104 and dbNSFP v4.1a to extract variant-level, amino acid-level, individuals’ outputs, and genome-level features. Additionally, protein sequences were generated using ENSP identifiers with the Ensembl API, and then encoded. The mutant sites’ ESM-1b and ProtTrans-T5 embeddings were subsequently extracted. Then, our model group (MMPatho) was developed by leveraging upon these efforts, which comprised ConsMM and EvoIndMM. To be specific, ConsMM employs individuals’ outputs and XGBoost with SHAP explanation analysis, while EvoIndMM investigates the potential enhancement of predictive capability by incorporating evolutionary information from ESM-1b and ProtT5-XL-U50, large protein language embeddings. Through rigorous comparative experiments, both ConsMM and EvoIndMM were capable of achieving remarkable AUROC (0.9836 and 0.9854) and AUPR (0.9852 and 0.9902) values on the blind test set devoid of overlapping variations and proteins from the training data, thus highlighting the superiority of our computational approach in the prediction of MM pathogenicity. Our Web server, available at http://csbio.njust.edu.cn/bioinf/mmpatho/, allows researchers to predict the pathogenicity (alongside the reliability index score) of MMs using the ConsMM and EvoIndMM models and provides extensive annotations for user input. Additionally, the newly constructed benchmark data set and blind test set can be accessed via the data page of our web server.

@article{doi101021acsjcim3c00950,
    author = "Ge, Fang and Arif, Muhammad and Yan, Zihao and Alahmadi, Hanin H. and Worachartcheewan, A. and Yu, Dong-Jun and Shoombuatong, Watshara",
    title = "MMPatho: Leveraging Multilevel Consensus and Evolutionary Information for Enhanced Missense Mutation Pathogenic Prediction",
    year = "2023",
    journal = "Journal of Chemical Information and Modeling",
    abstract = "Understanding the pathogenicity of missense mutation (MM) is essential for shed light on genetic diseases, gene functions, and individual variations. In this study, we propose a novel computational approach, called MMPatho, for enhancing missense mutation pathogenic prediction. First, we established a large-scale nonredundant MM benchmark data set based on the entire Ensembl database, complemented by a focused blind test set specifically for pathogenic GOF/LOF MM. Based on this data set, for each mutation, we utilized Ensembl VEP v104 and dbNSFP v4.1a to extract variant-level, amino acid-level, individuals’ outputs, and genome-level features. Additionally, protein sequences were generated using ENSP identifiers with the Ensembl API, and then encoded. The mutant sites’ ESM-1b and ProtTrans-T5 embeddings were subsequently extracted. Then, our model group (MMPatho) was developed by leveraging upon these efforts, which comprised ConsMM and EvoIndMM. To be specific, ConsMM employs individuals’ outputs and XGBoost with SHAP explanation analysis, while EvoIndMM investigates the potential enhancement of predictive capability by incorporating evolutionary information from ESM-1b and ProtT5-XL-U50, large protein language embeddings. Through rigorous comparative experiments, both ConsMM and EvoIndMM were capable of achieving remarkable AUROC (0.9836 and 0.9854) and AUPR (0.9852 and 0.9902) values on the blind test set devoid of overlapping variations and proteins from the training data, thus highlighting the superiority of our computational approach in the prediction of MM pathogenicity. Our Web server, available at http://csbio.njust.edu.cn/bioinf/mmpatho/, allows researchers to predict the pathogenicity (alongside the reliability index score) of MMs using the ConsMM and EvoIndMM models and provides extensive annotations for user input. Additionally, the newly constructed benchmark data set and blind test set can be accessed via the data page of our web server.",
    url = "https://doi.org/10.1021/acs.jcim.3c00950",
    doi = "10.1021/acs.jcim.3c00950",
    is_oa = "true",
    number = "22",
    pages = "7239-7257",
    semanticscholar_citation_count = "19",
    semanticscholar_id = "ee68df891e9599b35c7b9b6b98b17ad34fd18be9",
    volume = "63"
}

Machine learning models are increasingly used in time series prediction with promising results. The model explanation of time series prediction falls behind the model development and makes less sense to users in understanding model decisions. This paper proposes ES-Mask, a post-hoc and model-agnostic evolutionary strip mask-based saliency approach for time series applications. ES-Mask designs the mask consisting of strips with the same salient value in consecutive time steps to produce binary and sustained feature importance scores over time for easy understanding and interpretation of time series. ES-Mask uses an evolutionary algorithm to search for the optimal mask by manipulating strips in rounds, thus is agnostic to models by involving no internal model states in the search. The initial experiments on MIMIC-III data set show that ES-Mask outperforms state-of-the-art methods.

@inproceedings{doi101609aaaiv37i1327031,
    author = "Sun, Yifei and Song, Cheng and Lu, Feng and Li, Wei and Jin, Hai and Zomaya, A.",
    title = "ES-Mask: Evolutionary Strip Mask for Explaining Time Series Prediction (Student Abstract)",
    year = "2023",
    journal = "Proceedings of the AAAI Conference on Artificial Intelligence",
    booktitle = "AAAI Conference on Artificial Intelligence",
    abstract = "Machine learning models are increasingly used in time series prediction with promising results. The model explanation of time series prediction falls behind the model development and makes less sense to users in understanding model decisions. This paper proposes ES-Mask, a post-hoc and model-agnostic evolutionary strip mask-based saliency approach for time series applications. ES-Mask designs the mask consisting of strips with the same salient value in consecutive time steps to produce binary and sustained feature importance scores over time for easy understanding and interpretation of time series. ES-Mask uses an evolutionary algorithm to search for the optimal mask by manipulating strips in rounds, thus is agnostic to models by involving no internal model states in the search. The initial experiments on MIMIC-III data set show that ES-Mask outperforms state-of-the-art methods.",
    url = "https://ojs.aaai.org/index.php/AAAI/article/download/27031/26803",
    doi = "10.1609/aaai.v37i13.27031",
    is_oa = "true",
    number = "13",
    pages = "16342-16343",
    semanticscholar_citation_count = "1",
    semanticscholar_id = "a60f4b9173152f5cfde7fdea7b027444b81b4886",
    volume = "37"
}

In numerous contexts, ranging from systems safety assessment to finance and medical diagnosis, a relevant causal inference task is to predict unseen rare events—the so-called black swans. These are plausible, high-impact, but unexpected events for whose prediction a probabilistic-based causal inference falls short. For instance, a safety analyst needs to hypothesize potential rare co-causes that could lead to an accident, so as to manage the most unexpected failures besides the more obvious ones. Given an effect, we use abduction to support the generation of a plausible set of explanatory hypotheses for its causes. We present a generative evolutionary strategy—called Evolutionary Abduction (EVA)—for automating abductive inference by repeatedly constructing hypothetical cause-effect instances, and then automatically assessing their plausibility as well as their novelty with respect to already known instances—a mechanism mimicking the human reasoning employed whenever we need to select the best candidates from a set of hypotheses. Experiments with four datasets confirm that EVA can construct new and realistic multiple-cause hypotheses for a given effect. EVA outperforms alternative strategies based on probabilistic-based causal inference as well as state-of-the-art evolutionary algorithms, generating closer-to-real instances in most settings and datasets.

@article{doi1011453725855,
    author = "Pietrantuono, R. and Russo, S.",
    title = "Automatic Generation of Plausible Co-Occurring Causes for Effects Explanation or Prediction",
    year = "2025",
    journal = "ACM Transactions on Intelligent Systems and Technology",
    abstract = "In numerous contexts, ranging from systems safety assessment to finance and medical diagnosis, a relevant causal inference task is to predict unseen rare events—the so-called black swans. These are plausible, high-impact, but unexpected events for whose prediction a probabilistic-based causal inference falls short. For instance, a safety analyst needs to hypothesize potential rare co-causes that could lead to an accident, so as to manage the most unexpected failures besides the more obvious ones. Given an effect, we use abduction to support the generation of a plausible set of explanatory hypotheses for its causes. We present a generative evolutionary strategy—called Evolutionary Abduction (EVA)—for automating abductive inference by repeatedly constructing hypothetical cause-effect instances, and then automatically assessing their plausibility as well as their novelty with respect to already known instances—a mechanism mimicking the human reasoning employed whenever we need to select the best candidates from a set of hypotheses. Experiments with four datasets confirm that EVA can construct new and realistic multiple-cause hypotheses for a given effect. EVA outperforms alternative strategies based on probabilistic-based causal inference as well as state-of-the-art evolutionary algorithms, generating closer-to-real instances in most settings and datasets.",
    url = "https://www.semanticscholar.org/paper/8f41889324c31eb7751585c88340411560173083",
    doi = "10.1145/3725855",
    is_oa = "true",
    number = "3",
    pages = "1-31",
    semanticscholar_id = "8f41889324c31eb7751585c88340411560173083",
    volume = "16"
}

Advances in artificial intelligence (AI) and multi-omics integration are reshaping precision oncology by enabling deeper mechanistic understanding, improved characterization of tumor heterogeneity, and the accelerated discovery of targeted therapeutics. Antimicrobial peptides (AMPs) have emerged as promising candidates for cancer treatment due to their selective cytotoxicity, immunomodulatory properties, and ability to alter the tumor microenvironment. However, their accurate computational identification remains challenging because existing models struggle to capture the complex structural and functional determinants of AMP activity. In this study, we propose GAC-BiTCNN-AMP, a hybrid generative and explainable deep learning framework designed to advance peptide discovery for precision oncology. The architecture integrates a Generative Adversarial Network to enhance data diversity, Capsule Networks to model hierarchical molecular dependencies, and a Bidirectional Temporal Convolutional Neural Network for capturing contextual sequence information. To strengthen biological signal representation, the model incorporates embeddings from advanced protein language model including ProtTrans-T5, UniRep, and ESM-2 alongside a novel PsePSSM-DCT evolutionary descriptor. A wrapper-based XGBoost Forward Feature Selection strategy further refines the feature space by identifying the most discriminative sequence patterns. GAC-BiTCNN-AMP delivers strong predictive performance, achieving 97.42% accuracy and 0.923 MCC in cross-validation, and 95.32% accuracy with 0.914 MCC on the same independent test set. SHapley Additive exPlanations (SHAP) analysis highlights key contributions from the fused latent representations to peptide activity, demonstrating the framework's interpretability at the representation level. By integrating generative modeling, deep representation learning, and explainable AI, this study provides a scalable computational pipeline supporting therapeutic peptide discovery for targeted, immune-modulatory, and precision cancer applications.

@article{doi101038s41598026433706,
    author = "Ali, Farman and Khalid, Majdi and Alsini, Raed and Yafoz, Ayman and Alkhalifah, Tamim and Du, Meng-Ze",
    title = "A generative explainable model for antimicrobial peptide prediction using bidirectional temporal convolutional neural network.",
    year = "2026",
    journal = "Scientific reports",
    abstract = "Advances in artificial intelligence (AI) and multi-omics integration are reshaping precision oncology by enabling deeper mechanistic understanding, improved characterization of tumor heterogeneity, and the accelerated discovery of targeted therapeutics. Antimicrobial peptides (AMPs) have emerged as promising candidates for cancer treatment due to their selective cytotoxicity, immunomodulatory properties, and ability to alter the tumor microenvironment. However, their accurate computational identification remains challenging because existing models struggle to capture the complex structural and functional determinants of AMP activity. In this study, we propose GAC-BiTCNN-AMP, a hybrid generative and explainable deep learning framework designed to advance peptide discovery for precision oncology. The architecture integrates a Generative Adversarial Network to enhance data diversity, Capsule Networks to model hierarchical molecular dependencies, and a Bidirectional Temporal Convolutional Neural Network for capturing contextual sequence information. To strengthen biological signal representation, the model incorporates embeddings from advanced protein language model including ProtTrans-T5, UniRep, and ESM-2 alongside a novel PsePSSM-DCT evolutionary descriptor. A wrapper-based XGBoost Forward Feature Selection strategy further refines the feature space by identifying the most discriminative sequence patterns. GAC-BiTCNN-AMP delivers strong predictive performance, achieving 97.42\% accuracy and 0.923 MCC in cross-validation, and 95.32\% accuracy with 0.914 MCC on the same independent test set. SHapley Additive exPlanations (SHAP) analysis highlights key contributions from the fused latent representations to peptide activity, demonstrating the framework's interpretability at the representation level. By integrating generative modeling, deep representation learning, and explainable AI, this study provides a scalable computational pipeline supporting therapeutic peptide discovery for targeted, immune-modulatory, and precision cancer applications.",
    url = "https://pmc.ncbi.nlm.nih.gov/articles/PMC13128923/",
    doi = "10.1038/s41598-026-43370-6",
    pmcid = "PMC13128923",
    pmid = "41844874"
}

As a highly heterogeneous tumor, radiotherapy for Glioblastoma (GBM) is a complex and dynamic process. Traditional predictive methods of treatment response often rely on one or a few fixed time points, but this static approach may fail to capture the detailed, individualized changes occurring throughout the treatment process. To address these limitations, we proposed a novel approach called Evolutionary Heterogeneity Analysis Framework (EvoHAF), which integrates tumor heterogeneity and whole-process evolution of GBM radiotherapy. Our framework introduces an Image Heterogeneity Encoder, designed to capture the intricate spatial heterogeneity based on tumor subregions. Additionally, the Temporal Self-Attention Module (TSAM) mechanism integrates longitudinal imaging data throughout the course of radiotherapy, capturing the evolving nature of the tumor. We further introduce a Compensated Prediction Head (CPH) that dynamically refines predictions throughout the patient's radiotherapy. Experimental results on a cross-center cohort, including an internal dataset of 112 patients and an external validation dataset of 80 patients, demonstrate that EvoHAF achieves strong performance. For internal 5-fold validation, the AUC was 0.8519±0.0583, and for external validation, the AUC was 0.7675±0.0858. These results demonstrate the model's capability to provide accurate whole-process predictions. Moreover, the model's credibility is reinforced by providing visual explanations at both 2D and 3D subregional levels, establishing trust in its decisions and laying a strong foundation for clinical applications.

@article{doi101109jbhi20263668508,
    author = "Zheng, Yao and Huang, Dong and Wei, Jie and Liu, Tianci and Wu, Xiaoting and Feng, Yuefei and Chen, Chengwei and Liu, Yang",
    title = "Whole-Process Evolutionary Heterogeneity Analysis for Glioblastoma Radiotherapy Response Prediction.",
    year = "2026",
    journal = "IEEE journal of biomedical and health informatics",
    abstract = "As a highly heterogeneous tumor, radiotherapy for Glioblastoma (GBM) is a complex and dynamic process. Traditional predictive methods of treatment response often rely on one or a few fixed time points, but this static approach may fail to capture the detailed, individualized changes occurring throughout the treatment process. To address these limitations, we proposed a novel approach called Evolutionary Heterogeneity Analysis Framework (EvoHAF), which integrates tumor heterogeneity and whole-process evolution of GBM radiotherapy. Our framework introduces an Image Heterogeneity Encoder, designed to capture the intricate spatial heterogeneity based on tumor subregions. Additionally, the Temporal Self-Attention Module (TSAM) mechanism integrates longitudinal imaging data throughout the course of radiotherapy, capturing the evolving nature of the tumor. We further introduce a Compensated Prediction Head (CPH) that dynamically refines predictions throughout the patient's radiotherapy. Experimental results on a cross-center cohort, including an internal dataset of 112 patients and an external validation dataset of 80 patients, demonstrate that EvoHAF achieves strong performance. For internal 5-fold validation, the AUC was 0.8519±0.0583, and for external validation, the AUC was 0.7675±0.0858. These results demonstrate the model's capability to provide accurate whole-process predictions. Moreover, the model's credibility is reinforced by providing visual explanations at both 2D and 3D subregional levels, establishing trust in its decisions and laying a strong foundation for clinical applications.",
    url = "https://pubmed.ncbi.nlm.nih.gov/41770955/",
    doi = "10.1109/JBHI.2026.3668508",
    pmid = "41770955"
}

Anticancer peptides (ACPs) represent a promising class of therapeutic agents that selectively destroy cancer cells while sparing healthy tissues. Despite their potential, biological challenges-including poor biochemical stability, limited tumor selectivity, and inefficient delivery mechanisms - hinder their clinical translation. In parallel, the rapid expansion of peptide sequence data underscores the urgent need for accurate, scalable, and generalizable ACP prediction models. To address these limitations, we propose a robust hybrid deep learning framework-Mamba-ACP-that integrates transformer-based Evolutionary Scale Modeling (ESM-2) embeddings, handcrafted features (AAindex, BLOSUM62), and a Mamba-based sequence modeling architecture. This approach captures both evolutionary and physicochemical properties of peptides to enhance prediction performance. The model was trained and validated by using two benchmark datasets Set 1 and Set 2, commonly used in peptide-based computational biology. Mamba-ACP achieves 87.59% accuracy and an AUC of 0.9356 on Set 1, and 96.69% accuracy and an AUC of 0.9922 on Set 2, beating state-of-the-art ACP predictors like ACP-CapsPred and GRDF by a significant margin. The Mamba-ACP framework processes a token-level fused representation obtained by concatenating the token-level ESM-2 embeddings with the PCA-reduced handcrafted residue descriptors at each sequence position. These results affirm the effectiveness of combining pre-trained transformer embeddings with handcrafted features and structured sequence modeling in improving ACP classification. Our findings position Mamba-ACP as a new benchmark in computational peptide discovery, offering strong generalizability, reduced false positives, and efficient performance. We further provide model-level explanations via gradient-based residue/token saliency, SHAP feature importance for AAindex/BLOSUM62 descriptors, motif-level enrichment, and saliency-guided residue mutation validation.

@article{doi101109tcbbio20263684898,
    author = "Ashraf, Adeel and Pang, Yu-Shan and Jabbar, Abdul and Yu, Dong-Jun",
    title = "Mamba-ACP: a Hybrid State-Space and Transformer Framework for Interpretable Anticancer Peptide Prediction.",
    year = "2026",
    journal = "IEEE transactions on computational biology and bioinformatics",
    abstract = "Anticancer peptides (ACPs) represent a promising class of therapeutic agents that selectively destroy cancer cells while sparing healthy tissues. Despite their potential, biological challenges-including poor biochemical stability, limited tumor selectivity, and inefficient delivery mechanisms - hinder their clinical translation. In parallel, the rapid expansion of peptide sequence data underscores the urgent need for accurate, scalable, and generalizable ACP prediction models. To address these limitations, we propose a robust hybrid deep learning framework-Mamba-ACP-that integrates transformer-based Evolutionary Scale Modeling (ESM-2) embeddings, handcrafted features (AAindex, BLOSUM62), and a Mamba-based sequence modeling architecture. This approach captures both evolutionary and physicochemical properties of peptides to enhance prediction performance. The model was trained and validated by using two benchmark datasets Set 1 and Set 2, commonly used in peptide-based computational biology. Mamba-ACP achieves 87.59\% accuracy and an AUC of 0.9356 on Set 1, and 96.69\% accuracy and an AUC of 0.9922 on Set 2, beating state-of-the-art ACP predictors like ACP-CapsPred and GRDF by a significant margin. The Mamba-ACP framework processes a token-level fused representation obtained by concatenating the token-level ESM-2 embeddings with the PCA-reduced handcrafted residue descriptors at each sequence position. These results affirm the effectiveness of combining pre-trained transformer embeddings with handcrafted features and structured sequence modeling in improving ACP classification. Our findings position Mamba-ACP as a new benchmark in computational peptide discovery, offering strong generalizability, reduced false positives, and efficient performance. We further provide model-level explanations via gradient-based residue/token saliency, SHAP feature importance for AAindex/BLOSUM62 descriptors, motif-level enrichment, and saliency-guided residue mutation validation.",
    url = "https://pubmed.ncbi.nlm.nih.gov/41989903/",
    doi = "10.1109/TCBBIO.2026.3684898",
    pmid = "41989903"
}