SEARCH · PubMed Health

Results for “machine learning prediction”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 649 records · Page 36Linked to original sources

PubMed2025

ASGCL: Adaptive Sparse Mapping-based graph contrastive learning network for cancer drug response prediction.

Personalized cancer drug treatment is emerging as a frontier issue in modern medical research. Considering the genomic differences among cancer patients, determining the most effective drug treatment plan is a complex and crucial task. In response to these challenges, this study introduces the Adaptive Sparse Graph Contrastive Learning Network (ASGCL), an innovative approach to unraveling latent interactions in the complex context of cancer cell lines and drugs. The core of ASGCL is the GraphMorpher module, an innovative component that enhances the input graph structure via strategic node attribute masking and topological pruning. By contrasting the augmented graph with the original input, the model delineates distinct positive and negative sample sets at both node and graph levels. This dual-level contrastive approach significantly amplifies the model's discriminatory prowess in identifying nuanced drug responses. Leveraging a synergistic combination of supervised and contrastive loss, ASGCL accomplishes end-to-end learning of feature representations, substantially outperforming existing methodologies. Comprehensive ablation studies underscore the efficacy of each component, corroborating the model's robustness. Experimental evaluations further illuminate ASGCL's proficiency in predicting drug responses, offering a potent tool for guiding clinical decision-making in cancer therapy.

Humans↗

PubMed2006

Influence of amino acid properties for discriminating outer membrane proteins at better accuracy.

Discriminating outer membrane proteins (OMPs) from other folding types of globular and membrane proteins is an important task both for identifying outer membrane proteins from genomic sequences and for the successful prediction of their secondary and tertiary structures. In this work, we have analyzed the influence of physico-chemical, energetic and conformational properties of amino acid residues for discriminating outer membrane proteins using different machine learning algorithms, such as, Bayes rules, Logistic functions, Neural networks, Support vector machines, Decision trees, etc. We observed that most of the properties have discriminated the OMPs with similar accuracy. The neural network method with the property, free energy change could discriminate the OMPs from other folding types of globular and membrane proteins at the 5-fold cross-validation accuracy of 94.4% in a dataset of 1,088 proteins, which is better than that obtained with amino acid composition. The accuracy of discriminating globular proteins is 94.3% and that of transmembrane helical (TMH) proteins is 91.8%. Further, the neural network method is tested with globular proteins belonging to 30 major folding types and it could successfully exclude 99.4% of the considered 1612 non-redundant proteins. These accuracy levels are comparable to or better than other methods in the literature. We suggest that this method could be effectively used to discriminate OMPs and for detecting OMPs in genomic sequences.

Algorithms↗

PubMed2026

Artificial intelligence (AI) uses in stereotactic radiosurgery (SRS): diagnosis with brain metastasis (BM) - A systematic review.

BACKGROUND: Brain metastases (BM) are the most common intracranial tumors in adults, and stereotactic radiosurgery (SRS) has become a mainstay of management. However, several diagnostic challenges persist in the SRS pathway, particularly the differentiation of radiation necrosis (RN) from true tumor progression, which conventional MRI and even advanced imaging techniques often cannot reliably resolve. Recent advances in artificial intelligence (AI) offer the potential to address these diagnostic limitations. This systematic review synthesizes current literature on AI applications for MRI-based diagnostic decision support in BM patients undergoing SRS, with a focus on radiomics and deep learning tools for distinguishing RN from progression, classifying molecular and histologic subtypes, and predicting treatment response. METHODS: A systematic review was performed in accordance with PRISMA guidelines. PubMed, Web of Science, and Scopus were searched using a targeted query combining terms related to AI, brain metastasis, diagnosis or imaging, and SRS. After screening 483 records and applying strict inclusion and exclusion criteria, 18 studies published between 2015 and 2025 were included. Data were extracted on study design, cohort characteristics, imaging modality, AI methodology, validation strategy, and reported diagnostic performance. RESULTS: Among the 18 included studies, AI models demonstrated strong performance across diagnostic tasks in the BM-SRS pathway. The differentiation of RN from true tumor progression was the most extensively studied application, addressed by 14 of 18 studies, with reported AUCs ranging from 0.71 to 0.94. Support vector machines, random-forest ensembles, convolutional neural networks, and transformer-based multimodal architectures were widely used. The literature evolved from single-sequence radiomic classifiers in 2018 to multimodal deep learning frameworks fusing imaging with clinical and genomic data in 2025. Contrast-enhanced T1-weighted MRI was the dominant imaging input, and texture-based radiomic features (GLCM, GLSZM, GLDM, and wavelet-derived features) were the most consistently predictive. The highest-performing models reached AUCs of 0.85-0.91 through multimodal integration of imaging with clinical and genomic features, and consistently outperformed expert neuroradiologist read on matched cases. Remaining studies addressed longitudinal segmentation-based detection of local failure and adverse radiation effects, BRAF mutation status in melanoma BM, early Gamma Knife treatment response, and primary tumor histology classification, with more variable performance. CONCLUSION: AI models, particularly those integrating MRI-derived radiomic features with clinical and genomic data, show high accuracy in supporting diagnostic decisions for BM patients treated with SRS. The post-SRS differentiation of radiation necrosis from true tumor progression has reached the greatest level of maturity and is closest to clinical translation, with potential to reduce unnecessary biopsies, personalize surveillance intervals, and rationalize treatment-pathway decisions. Other diagnostic applications, including molecular subtyping and primary tumor histology classification, remain exploratory and require further multicenter validation. Integration of AI tools into multidisciplinary tumor-board workflows, combined with prospective validation and standardized reporting, will be essential to realize the full clinical benefits of AI in SRS for brain metastases.

Humans↗

PubMed2026

Listening forward: emerging roles of bioacoustics in ecology, evolution, and conservation.

Bioacoustics is increasingly shifting from a mostly descriptive pursuit to one that can anticipate ecological change. Recent innovations-from autonomous recording units and edge-computing sensors to speech-inspired feature extraction and machine-learning techniques like transfer learning, unsupervised discovery, and explainable AI-are transforming the study of animal communication. These advances let us work at scales previously difficult to imagine. Automated species recognition, individual identification, and even tracking cultural evolution over decades are now within reach. Entire ecosystem soundscapes can be mapped with unprecedented resolution. Looking ahead, global listening networks, adaptive acoustic indices, and live biodiversity dashboards seem increasingly realistic. We may soon build digital models that simulate communication networks under future scenarios. Closer integration with genomics, physiology, and robotics could link vocal traits to their genetic, physiological, and ecological drivers. Challenges remain, including data governance, acoustic privacy, and equitable access to the planet's sonic heritage. Bioacoustics may be on the way to becoming a predictive, integrative science - one particularly well suited to monitoring, interpreting, and helping safeguard life's communication systems in a rapidly changing world.

Animals↗

PubMed2026

Integrating histology and spatial transcriptomics via multimodal transformers and contrastive representation learning for accurate gene expression prediction.

Predicting spatial gene expression from Histological images is a fundamental task in understanding tissue organization and molecular phenotypes. However, existing methods often rely on single-model representations or lack effective alignment between image and transcriptomic features. To address these limitations, we propose a unified multimodal learning framework that integrates histological imaging and spatial transcriptomics through a shared latent representation space. Specifically, histological H&E images are encoded by a ResNet50-based convolutional stem and a MobileViT Transformer backbone to extract hierarchical visual representations. Both modalities are projected into a shared latent space via linear-GELU-dropout transformation blocks, enabling cross-modal alignment through a contrastive learning objective that maximizes agreement between the corresponding image and the spot embeddings. Experimental results on the 10x Genomics Visium dataset of human liver tissue demonstrate that MViTGene achieves significantly higher prediction accuracy than existing methods across multiple gene subsets, with improvements of 20%, 33%, and 12% in predicting marker genes, highly expressed genes, and highly variable genes, respectively. The significant improvement in relevance indicates that the model can more accurately capture the true correspondence between tissue morphology and gene expression, therefore enabling more reliable biological interpretation. It provides a computational tool for high-throughput spatial gene expression prediction that balances performance and interpretability.

Humans↗

PubMed2000

[Value of combined detection of tumor markers for the prediction of small cell and non-small cell lung cancer].

To evaluate the value of detection of 4 tumor markers(CEA, CA125, gastrin, and NSE) for histological types in patients with lung cancer and to improve the predicted efficiency of tumor markers for distinguishing between small cell lung cancer(SCLC) and non-small cell lung cancer (NSCLC), these 4 tumor markers in serum were determined in 51 patients (21 cases with SCLC, 30 cases with NSCLC) with confirmed primary diagnosis of lung cancer of different histology by radioimmunoassay. Linear learning machine method, PRIMA method and KNN method were used to classify SCLC and NSCLC. The levels of gastrin and NSE in SCLC were apparently higher than those of gastrin and NSE in NSCLC, but the levels of CEA and CA125 in SCLC were significantly lower than those in NSCLC. Smoking had an effect on the levels of CEA and CA125, but had little effect on those of gastrin and NSE. The total accuracy of the three methods was over 85% in distinguishing SCLC from NSCLC. So combined detection of the four tumor markers in serum might be useful in the prediction of histological types in patients with lung cancer.

Adenocarcinoma↗

PubMed2006

Structural bioinformatics prediction of membrane-binding proteins.

Membrane-binding peripheral proteins play important roles in many biological processes, including cell signaling and membrane trafficking. Unlike integral membrane proteins, these proteins bind the membrane mostly in a reversible manner. Since peripheral proteins do not have canonical transmembrane segments, it is difficult to identify them from their amino acid sequences. As a first step toward genome-scale identification of membrane-binding peripheral proteins, we built a kernel-based machine learning protocol. Key features of known membrane-binding proteins, including electrostatic properties and amino acid composition, were calculated from their amino acid sequences and tertiary structures, which were then incorporated into the support vector machine to perform the classification. A data set of 40 membrane-binding proteins and 230 non-membrane-binding proteins was used to construct and validate the protocol. Cross-validation and holdout evaluation of the protocol showed that the accuracy of the prediction reached up to 93.7% and 91.6%, respectively. The protocol was applied to the prediction of membrane-binding properties of four C2 domains from novel protein kinases C. Although these C2 domains have 50% sequence identity, only one of them was predicted to bind the membrane, which was verified experimentally with surface plasmon resonance analysis. These results suggest that our protocol can be used for predicting membrane-binding properties of a wide variety of modular domains and may be further extended to genome-scale identification of membrane-binding peripheral proteins.

Amino Acid Sequence↗

PubMed2025

Establishment of a prognostic model based on ER stress-related cell death genes and proposing a novel combination therapy in acute myeloid leukemia.

BACKGROUND: Acute myeloid leukemia (AML) is a highly heterogeneous malignancy, presenting significant challenges in accurately predicting patient prognosis. Dysregulation of endoplasmic reticulum (ER) stress and resistance to programmed cell death (PCD) are hallmarks of AML cells. However, the prognostic significance of the interplay between ER stress and cell death pathways in AML remains largely unexplored. METHODS: We analyzed RNA sequencing and clinical data from 887 AML patients across 4 cohorts to develop an ER stress-related cell death index (ERCDI) using 10 machine-learning algorithms with 117 unique combinations. Survival and time-dependent Receiver Operating Characteristic Curve (ROC) analyses were performed to assess the model's efficacy. Clinical characteristics, the tumor immune microenvironment, and drug sensitivity differences between the high- and low-risk groups were also analyzed. The CMap database was used to identify potential therapeutic drugs. In vitro and in vivo experiments, including CCK-8, colony formation, flow cytometry, Transwell assays, and xenograft mouse models, were conducted to evaluate the effects of the target genes and candidate drugs. RESULTS: The ERCDI demonstrated strong prognostic and predictive performance for prognosis in AML patients. Furthermore, the ERCDI effectively predicted immunotherapy and chemotherapy outcomes and was associated with the immune features of the different risk groups. DNA damage-inducible transcript 4 protein (DDIT4), a key gene associated with ERCDI, is related to poor prognosis in AML patients with high expression. Additionally, the knockdown of DDIT4 significantly inhibited AML cell proliferation, induced cell apoptosis, and promoted cell cycle arrest. Chaetocin was subsequently identified as a candidate compound for AML treatment. Subsequent experiments suggested that combining chaetocin and venetoclax is a potentially promising therapeutic strategy for AML. CONCLUSION: The ERCDI provides personalized risk assessment and treatment recommendations for individual AML patients. The combined use of chaetocin and venetoclax can potentially be repurposed for AML therapy.

Humans↗

PubMedDate not supplied

Effect of selection of molecular descriptors on the prediction of blood-brain barrier penetrating and nonpenetrating agents by statistical learning methods.

The ability or inability of a drug to penetrate into the brain is a key consideration in drug design. Drugs for treating central nervous system (CNS) disorders need to be able to penetrate the blood-brain barrier (BBB). BBB nonpenetration is desirable for non-CNS-targeting drugs to minimize potential CNS-related side effects. Computational methods have been employed for the prediction of BBB-penetrating (BBB+) and -nonpenetrating (BBB-) agents at impressive accuracies of 75-92% and 60-80%, respectively. However, the majority of these studies give a substantially lower BBB- accuracy, and thus overall accuracy, than the BBB+ accuracy. This work examined whether proper selection of molecular descriptors can improve both the BBB- and the overall accuracies of statistical learning methods. The methods tested include logistic regression, linear discriminate analysis, k nearest neighbor, C4.5 decision tree, probabilistic neural network, and support vector machine. Molecular descriptors were selected by using a feature selection method, recursive feature elimination (RFE). Results by using 415 BBB+ and BBB- agents show that RFE substantially improves both the BBB- and the overall accuracy for all of the methods studied. This suggests that statistical learning methods combined with proper feature selection is potentially useful for facilitating a more balanced and improved prediction of BBB+ and BBB- agents.

Artificial Intelligence↗

PubMed2026

Cooccurrence of Homologous Recombination Deficiency and Mismatch Repair Deficiency in Colorectal Cancer.

Homologous recombination deficiency (HRD) in colorectal cancer (CRC) remains largely unexplored. In contrast, mismatch repair deficiency (dMMR) occurs in ∼15% of patients with CRC. Although HRD and dMMR have historically been regarded as mutually exclusive, emerging evidence suggests that this mutual exclusivity may not be absolute. Here, we conducted a retrospective cohort study utilizing genomic and transcriptomic data to define HRD status in a Chinese dMMR CRC cohort (n = 99). Multiple machine learning approaches were employed to analyze the expression profiles of these tumors and to develop a classifier distinguishing HRD from homologous recombination proficiency (HRP) in dMMR CRCs. In the Chinese dMMR CRC cohort, 66% of tumors were classified as HRD. Compared with the HRP group, the HRD group had a significantly higher tumor mutational burden and better outcomes. The derived expression signature, comprising eight genes, successfully predicted HRD status in dMMR tumors with high accuracy in the training set (AUC = 0.88, Naïve Bayes) and the test set (AUC = 0.87). In this study, a subset of dMMR CRC tumors with co-occurring HRD was identified, which may have potential implications for patient stratification and the application of targeted therapies, such as PARP inhibitors, in this molecular subgroup.

colorectal cancer↗

PubMed2026

Sex Hormone Receptors, HBV Integrations and Their Prognostic Predictive Value Among Hepatocellular Carcinoma Patients.

Hepatocellular carcinoma (HCC) related to hepatitis B virus (HBV) infection predominantly affects males, yet few studies have investigated the association between sex hormones and HBV integrations, and their involvement in HCC prognosis. We assessed estrogen receptor alpha (ERα) and androgen receptor (AR) expression via immunohistochemistry on tissue microarrays constructed from 426 HBV-related HCC samples. HBV integration features were determined using HBV-captured sequencing data. Logistic regression models were utilized to evaluate the association between sex hormone receptor expression level and HBV integration features. Cox regression models, combined with machine learning (ML) methods, were implemented to investigate the prognostic value of sex hormone receptors and HBV integrations concerning overall survival. We found high AR expression level was significantly associated with higher HBV integration levels (adjusted odds ratio [aOR] = 1.84, 95% confidence interval [CI]: 1.09-3.11, P for trend = 0.012), TERT integration (aOR = 2.34, 95% CI: 1.16-4.74, P for trend = 0.047), intergenic integration (aOR = 2.25, 95% CI: 1.20-4.24, P for trend = 0.021), and promoter integration (aOR = 1.81, 95% CI: 1.00-3.31, P for trend = 0.034). The inclusion of sex hormone receptors and HBV integrations in the predictive models led to improvements across all performance metrics in the Cox regression analyses (AUC improvement: 0.014 [Training], 0.026 [Validation]) and the ML (AUC improvement: 0.022 [Training]), although a slight deterioration in performance was noted in the ML validation set. The results suggested a relationship between AR expression level and HBV integration events, as well as the potential utility of HBV integration biomarkers and sex hormone receptor profiles in assessing post-surgical prognosis among HCC patients.

Humans↗

PubMed2026

Cross-Platform Proteomics and Machine Learning Algorithms Nominate Plasma Biomarkers of Stroke Diagnosis.

BACKGROUND: Blood-based biomarkers for stroke subtyping could improve triage in emergency settings. We used cross-platform proteomics to identify plasma biomarkers differentiating major stroke diagnostic groups. METHODS: We conducted a case-control study using 2 biorepositories. Plasma was collected in the emergency department from adults with suspected stroke before therapeutic intervention. Differentially enriched proteins were identified across acute ischemic stroke, intracerebral hemorrhage, transient ischemic attack, and stroke mimics using SomaScan discovery proteomics (Grady). Differentially enriched proteins were nominated using pairwise and multigroup comparisons and adjusted for clinical covariates. Protein panels were created using least absolute shrinkage and selection operator logistic regression. Internal validation used repeated nested cross-validation (rCV) and targeted mass spectrometry (MS), while external validation used data-independent acquisition  mass spectrometry in an independent cohort (Yale). RESULTS: We included 100 subjects (40 with acute ischemic stroke, 20 with intracerebral hemorrhage, 20 with transient ischemic attack, 20 with stroke mimics) in discovery and 80 subjects (20 per group) in external validation cohorts. SomaScan quantified 7307 proteins, of which 61 differentiated stroke subtypes. We identified 7 protein classifiers for acute ischemic stroke (rCV-area under the curve, 0.82 [95% CI, 0.78-0.86]), 6 for intracerebral hemorrhage (rCV-area under the curve, 0.70 [95% CI, 0.64-0.76]), 8 for transient ischemic attack (rCV-area under the curve, 0.78 [95% CI, 0.73-0.84]), and 7 for stroke mimics (rCV-area under the curve, 0.81 [95% CI, 0.77-0.86]). Targeted proteomics internally validated 11 proteins, and data-independent acquisition-mass spectrometry externally validated 32 proteins, including VTN (vitronectin), PLG (plasminogen), and S100A9 as top stroke mimics, transient ischemic attack, and intracerebral hemorrhage classifiers. CONCLUSIONS: This study highlights plasma proteomics as a valuable tool for discovering protein biomarkers of stroke diagnosis. These findings support further validation in larger, multicenter cohorts to facilitate biomarker-guided stroke diagnosis in acute care.

Humans↗

PubMed2025

A SuperLearner-based pipeline for the development of DNA methylation-derived predictors of phenotypic traits.

BACKGROUND: DNA methylation (DNAm) provides a window to characterize the impacts of environmental exposures and the biological aging process. Epigenetic clocks are often trained on DNAm using penalized regression of CpG sites, but recent evidence suggests potential benefits of training epigenetic predictors on principal components. METHODOLOGY/FINDINGS: We developed a pipeline to simultaneously train three epigenetic predictors; a traditional CpG Clock, a PCA Clock, and a SuperLearner PCA Clock (SL PCA). We gathered publicly available DNAm datasets to generate i) a novel childhood epigenetic clock, ii) a reconstructed Hannum adult blood clock, and iii) as a proof of concept, a predictor of polybrominated biphenyl exposure using the three developmental methodologies. We used correlation coefficients and median absolute error to assess fit between predicted and observed measures, as well as agreement between duplicates. The SL PCA clocks improved fit with observed phenotypes relative to the PCA clocks or CpG clocks across several datasets. We found evidence for higher agreement between duplicate samples run on alternate DNAm arrays when using SL PCA clocks relative to traditional methods. Analyses examining associations between relevant exposures and epigenetic age acceleration (EAA) produced more precise effect estimates when using predictions derived from SL PCA clocks. CONCLUSIONS: We introduce a novel method for the development of DNAm-based predictors that combines the improved reliability conferred by training on principal components with advanced ensemble-based machine learning. Coupling SuperLearner with PCA in the predictor development process may be especially relevant for studies with longitudinal designs utilizing multiple array types, as well as for the development of predictors of more complex phenotypic traits.

DNA Methylation↗

PubMed2026

The AI Revolution: Shaping the Present and Future of Pharmaceutical Research and Development.

The transformative role of artificial intelligence (AI) in the pharmaceutical industry is examined, with a focus on its significant contributions to drug discovery, development, and clinical trial processes. It highlights the inefficiencies and high costs associated with traditional drug development and explores how AI and machine learning (ML) can enhance these processes by analyzing extensive biological datasets. The historical context of AI in pharmaceutical development is examined, noting how advances in computational power and data accessibility have facilitated innovative methodologies, such as predictive analytics and natural language processing. Contemporary trends reveal the integration of AI technologies in drug design, repurposing, and patient response forecasting. This study also addresses the challenges of participant recruitment for clinical trials and proposes AI-driven solutions to optimize patient selection and data management. Furthermore, it discusses AI's role in tailored medicine, emphasizing its potential for advancing precision therapy through targeted drug development and personalized treatment strategies. The importance of digital tools, genomic data analysis, and AI-driven imaging technologies for customizing therapeutic approaches is underscored, along with the regulatory and ethical challenges posed by AI deployment in healthcare. This study illustrates the complexities of AI applications in the pharmaceutical sector, offering insights into both successful and unsuccessful initiatives. The findings suggest that the digitalization of the pharmaceutical industry and enhanced AI integration hold promise for developing safer and more effective therapeutic strategies, while also identifying obstacles to their widespread adoption and optimal functionality.

Artificial intelligence↗

PubMed1998

Using classification tree and logistic regression methods to diagnose myocardial infarction.

Early and accurate diagnosis of myocardial infarction (MI) in patients who present to the Emergency Room (ER) complaining of chest pain is an important problem in emergency medicine. A number of decision aids have been developed to assist with this problem but have not achieved general use. Machine learning techniques, including classification tree and logistic regression (LR) methods, have the potential to create simple but accurate decision aids. Both a classification tree (FT Tree) and an LR model (FT LR) have been developed to predict the probability that a patient with chest pain is having an MI based solely upon data available at time of presentation to the ER. Training data came from a data set collected in Edinburgh, Scotland. Each model was then tested on a separate Edinburgh data set, as well as on a data set from a different hospital in Sheffield, England. Previously published models, the Goldman classification tree[1] and Kennedy LR equation[2], were evaluated on the same test data sets. On the Edinburgh test set, results showed that the FT Tree, FT LR, and Kennedy LR performed equally well, with ROC curve areas of 94.04%, 94.28%, and 94.30%, respectively, while the Goldman Tree's performance was significantly poorer, with an area of 84.03%. The difference in ROC areas between the first three models and the Goldman model is significant beyond the 0.0001 level. On the Sheffield test set, results showed that the FT Tree, FT LR, and Kennedy LR ROC areas were not significantly different (p > = 0.17), while the FT Tree again outperformed the Goldman Tree (p = 0.006). Unlike previous work[3], this study indicates that classification trees, which have certain advantages over LR models, may perform as well as LR models in the diagnosis of patients with MI.

Algorithms↗

PubMed2005

PCP: a program for supervised classification of gene expression profiles.

UNLABELLED: PCP (Pattern Classification Program) is an open-source machine learning program for supervised classification of patterns (vectors of measurements). The principal use of PCP in bioinformatics is design and evaluation of classifiers for use in clinical diagnostic tests based on measurements of gene expression. PCP implements leading pattern classification and gene selection algorithms and incorporates cross-validation estimation of classifier performance. Importantly, the implementation integrates gene selection and class prediction stages, which is vital for computing reliable performance estimates in small-sample scenarios. Additionally, the program includes automated and efficient model selection (optimization of parameters) for support vector machine (SVM) classifier. The distribution includes Linux and Windows/Cygwin binaries. The program can easily be ported to other platforms. AVAILABILITY: Free download at http://pcp.sourceforge.net

Algorithms↗

PubMed2006

Analysis and prediction of helix shift errors in homology modeling.

High sequence identity between two proteins (e.g. > 60%) is a strong evidence for high structural similarity. However, internal shifts in one of the two proteins can sometimes give rise to unexpectedly high structural differences. This, in turn, causes unreliable structure predictions when two such proteins are used in homology modeling. Here, we perform a computational analysis of helix shifts and we show that their occurrence can be predicted with statistical learning methods. Our results indicate that helix shifts increase the RMS error by factor 2.6 compared to those protein pairs without a helix shift. Although helix shifts are rare (1.6% of helices and a commensurately higher number of proteins are affected), they therefore pose a significant problem for reliable structure prediction systems. In this paper, we prototype a new approach for model quality assessment and demonstrate that it can successfully warn against helix shifts. A support vector machine trained on a wide range of sequence and structure properties predicts the occurrence of helix shifts with a sensitivity of 74.2% and a specificity of 83.6%. On an equalized test dataset, this corresponds to an accuracy of 78.9%. Projected to the full dataset, it translates to an accuracy of 83.4%. Our analysis shows that helix shift detection is a valuable building block for highly reliable structure prediction systems. Furthermore, the statistical learning based approach to helix shift detection that we employ here is orthogonal to well-established model quality assessment methods (which use geometric constraint checking or mean force potentials). Therefore, a further increase of prediction accuracy is expected from the combination of these methods.

Computational Biology↗

PubMed2026

Topologically distinct intratumoral heterogeneity scores for predicting high-risk pathological grades in invasive lung adenocarcinoma: A multicenter study across four institutions.

High-risk subtypes of invasive lung adenocarcinoma (IAC), particularly micropapillary- or solid-predominant patterns, are closely associated with poor prognosis. This multicenter retrospective study developed and validated a predictive model for the preoperative identification of these high-risk subtypes using topologically distinct intratumoral heterogeneity (ITH) scores derived from CT images. The study included 1,051 patients with IAC. Two complementary ITH scores were developed: a two-dimensional ITH score, which integrated local radiomics features with global pixel distribution patterns on the largest cross-sectional CT slice, and a three-dimensional ITH score, which extended this quantification across the entire tumor volume. Clinicoradiological features and ITH scores were incorporated as model inputs to construct six base machine learning classifiers and a final stacking ensemble classifier. Model interpretability and robustness were evaluated using SHapley Additive exPlanations (SHAP)-based ablation analyses. An independent dataset from The Cancer Imaging Archive (TCIA) was used for external validation to investigate associations between ITH scores and pathological characteristics, genomic features, recurrence-free survival, and overall survival. The stacking ensemble classifier achieved the best predictive performance, with an area under the receiver operating characteristic curve of 0.875, outperforming models based solely on radiomics features (0.834) or clinicoradiological features (0.792). SHAP analysis identified the 3D ITH score as the most influential contributor to model output, and TCIA validation showed that higher 3D ITH scores were associated with more aggressive tumor biology and poorer survival outcomes. The topologically distinct 3D ITH score may provide a clinically meaningful imaging biomarker for preoperative risk stratification in IAC.

Journal Article↗