Anticancer peptides (ACPs) are bioactive compounds known for their selective cytotoxicity against tumor cells via various mechanisms. Recent studies have demonstrated that in silico machine learning methods are effective in predicting peptides with anticancer activity. In this study, we collected and analyzed over a thousand experimentally verified ACPs, specifically targeting peptides derived from natural sources. We developed a precise prediction model based on their sequence and structural features, and the model’s evaluation results suggest its strong predictive ability for anticancer activity. To enhance reliability, we integrated the results of this model with those from other available methods. In total, we identified 176 potential ACPs, some of which were synthesized and further evaluated using the MTT colorimetric assay. All of these putative ACPs exhibited significant anticancer effects and selective cytotoxicity against specific tumor cells. In summary, we present a strategy for identifying and characterizing natural peptides with selective cytotoxicity against cancer cells, which could serve as novel therapeutic agents. Our prediction model can effectively screen new molecules for potential anticancer activity, and the results from in vitro experiments provide compelling evidence of the candidates’ anticancer effects and selective cytotoxicity.

Background: Short peptide epitopes are valuable for mechanistic studies, yet their intrinsic low immunogenicity and lack of commercial antibodies hinder rapid antibody generation. Methods: We developed a reproducible, sequence-level workflow combining cross-species/structural triage, independent MHC-I/II prioritization, and conservative heteroclitic-style substitutions to enhance predicted MHC affinity while preserving native epitope features. Using visfatin as a model, two optimized fragments were conjugated to KLH and tested in mice for antibody titers, isotype profiles, and binding kinetics. Results: Mutant peptides improved MHC-binding prediction, elicited stronger antibody titers, and promoted isotype maturation (increased IgG1). Importantly, antibodies maintained measurable binding to wild-type sequences, indicating preserved cross-recognition. Similar effects were reproduced with additional antigens. Conclusions: This proof-of-concept study, based on small exploratory mouse cohorts (n = 3 per group), demonstrates that strategic, minimal sequence edits can significantly enhance peptide immunogenicity while preserving native epitope recognition. This streamlined workflow provides a low-barrier route to generate epitope-directed antibodies when commercial reagents are unavailable.

Identifying the tumor-associated antigens (TAAs) overexpressed in a subgroup of tumor patients is a substantial challenge for cancer treatment. Although there are several methods based on the concept of differential expression, there is a lack of proper algorithms based on the heterogeneous transcriptome expression for exploring effective TAAs. Here, we propose an algorithm, TAPINTO, to objectively predict overexpressed TAAs whose expression is heterogeneous in cancer patients. This algorithm exploits the dispersion of expression in a subgroup of patients to create 3 quantitative parameters (the specific average expression, frequency, and fold change) for evaluating potential TAAs and has a good performance compared with other approaches. Based on these parameters, TAPINTO successfully identified HER2, a famous therapeutic target, and other potential TAAs (CXCL9, KCNJ3, SQLE, MMP11, and SLC7A2) in breast cancer; moreover, these parameters were dramatically consistent with the trend of clinical outcomes (objective response rate, progression-free survival, and serious adverse effects) of therapeutic antibodies. The ability of TAPINTO to capture heterogeneous expression patterns among patients was further validated in cancer hallmarks, subtypes, and prognosis. This study suggests that this novel method will enable potential TAAs to facilitate the subgroup of patients for diagnosis, prognostication, and therapy to overcome the tumor heterogeneity.

Hepatitis C virus (HCV) infection remains a significant global health burden, driven by the emergence of drug-resistant strains and the limited efficacy of current antiviral therapies. A promising strategy for therapeutic intervention involves targeting the NS3 protease, a viral enzyme essential for replication. In this study, we present the first computational model specifically designed to identify NS3 protease inhibitory peptides (NS3IPs). Using amino acid composition (AAC) and K-spaced amino acid pair composition (CKSAAP) features, we developed machine learning classifiers based on support vector machine (SVM) and random forest (RF), achieving accuracies of 98.85% and 97.83%, respectively, validated through 5-fold cross-validation and independent testing. To support the accessibility of the strategy, we implemented a web-based tool, iDNS3IP, which enables real-time prediction of NS3IPs. In addition, we performed feature space analyses using PCA, t-SNE, and LDA based on AAindex descriptors. The resulting visualizations showed a distinguishable clustering between NS3IPs and non-inhibitory peptides, suggesting that inhibitory activity may correlate with characteristic physicochemical patterns. This study provides a reliable and interpretable platform to assist in the discovery of therapeutic peptides and supports continued research into peptide-based antiviral strategies for drug-resistant HCV. To enhance its flexibility, the iDNS3IP web tool also incorporates a BLAST-based similarity search function, enabling users to evaluate inhibitory candidates from both predictive and homology-based perspectives.

This study aimed to develop and validate a transcriptomic risk signature for uterine corpus endometrial carcinoma (UCEC) and to investigate whether the identified prognostic program reflects immune–tumor uncoupling within the tumor microenvironment. Using transcriptomic data from The Cancer Genome Atlas (TCGA) UCEC cohort, we identified a 28-gene transcriptomic signature defining a high-risk state. The derived risk score robustly stratified patients into distinct survival groups and remained an independent predictor of overall survival after adjustment for clinical covariates. Functional analyses revealed that high-risk tumors are characterized by a distinct immune–tumor uncoupling phenotype, in which interferon-gamma (IFNG)-associated inflammatory signaling is preserved but fails to translate into effective antitumor immune activity. Specifically, effector immune programs, including CD8 T cell-related signatures and cytotoxic activity, were consistently reduced despite elevated IFNG-associated signaling, indicating a functional discordance between immune activation and immune execution rather than classical T cell exhaustion. In parallel, high-risk tumors exhibited consistently elevated cell cycle and DNA repair-associated transcriptional programs, suggesting that proliferative and stress-adaptive mechanisms represent dominant drivers of poor prognosis. External assessment in an independent GEO cohort (GSE17025) demonstrated consistent associations between signature activity, tumor status, and histological grade, supporting the reproducibility of the underlying transcriptional program at the biological and clinicopathological level. Collectively, this study provides transcriptomic evidence for a previously underappreciated immune–tumor uncoupling state in UCEC and highlights the importance of integrating immune signaling and tumor-intrinsic programs to understand disease progression.