RESUMO
PURPOSE: Rare cancers constitute over 20% of human neoplasms, often affecting patients with unmet medical needs. The development of effective classification and prognostication systems is crucial to improve the decision-making process and drive innovative treatment strategies. We have created and implemented MOSAIC, an artificial intelligence (AI)-based framework designed for multimodal analysis, classification, and personalized prognostic assessment in rare cancers. Clinical validation was performed on myelodysplastic syndrome (MDS), a rare hematologic cancer with clinical and genomic heterogeneities. METHODS: We analyzed 4,427 patients with MDS divided into training and validation cohorts. Deep learning methods were applied to integrate and impute clinical/genomic features. Clustering was performed by combining Uniform Manifold Approximation and Projection for Dimension Reduction + Hierarchical Density-Based Spatial Clustering of Applications with Noise (UMAP + HDBSCAN) methods, compared with the conventional Hierarchical Dirichlet Process (HDP). Linear and AI-based nonlinear approaches were compared for survival prediction. Explainable AI (Shapley Additive Explanations approach [SHAP]) and federated learning were used to improve the interpretation and the performance of the clinical models, integrating them into distributed infrastructure. RESULTS: UMAP + HDBSCAN clustering obtained a more granular patient stratification, achieving a higher average silhouette coefficient (0.16) with respect to HDP (0.01) and higher balanced accuracy in cluster classification by Random Forest (92.7% ± 1.3% and 85.8% ± 0.8%). AI methods for survival prediction outperform conventional statistical techniques and the reference prognostic tool for MDS. Nonlinear Gradient Boosting Survival stands in the internal (Concordance-Index [C-Index], 0.77; SD, 0.01) and external validation (C-Index, 0.74; SD, 0.02). SHAP analysis revealed that similar features drove patients' subgroups and outcomes in both training and validation cohorts. Federated implementation improved the accuracy of developed models. CONCLUSION: MOSAIC provides an explainable and robust framework to optimize classification and prognostic assessment of rare cancers. AI-based approaches demonstrated superior accuracy in capturing genomic similarities and providing individual prognostic information compared with conventional statistical methods. Its federated implementation ensures broad clinical application, guaranteeing high performance and data protection.
Assuntos
Inteligência Artificial , Medicina de Precisão , Humanos , Prognóstico , Medicina de Precisão/métodos , Feminino , Doenças Raras/classificação , Doenças Raras/genética , Doenças Raras/diagnóstico , Masculino , Aprendizado Profundo , Neoplasias/classificação , Neoplasias/genética , Neoplasias/diagnóstico , Síndromes Mielodisplásicas/diagnóstico , Síndromes Mielodisplásicas/classificação , Síndromes Mielodisplásicas/genética , Síndromes Mielodisplásicas/terapia , Algoritmos , Pessoa de Meia-Idade , Idoso , Análise por ConglomeradosRESUMO
MOTIVATION: Mutational signatures are a critical component in deciphering the genetic alterations that underlie cancer development and have become a valuable resource to understand the genomic changes during tumorigenesis. Therefore, it is essential to employ precise and accurate methods for their extraction to ensure that the underlying patterns are reliably identified and can be effectively utilized in new strategies for diagnosis, prognosis, and treatment of cancer patients. RESULTS: We present MUSE-XAE, a novel method for mutational signature extraction from cancer genomes using an explainable autoencoder. Our approach employs a hybrid architecture consisting of a nonlinear encoder that can capture nonlinear interactions among features, and a linear decoder which ensures the interpretability of the active signatures. We evaluated and compared MUSE-XAE with other available tools on both synthetic and real cancer datasets and demonstrated that it achieves superior performance in terms of precision and sensitivity in recovering mutational signature profiles. MUSE-XAE extracts highly discriminative mutational signature profiles by enhancing the classification of primary tumour types and subtypes in real world settings. This approach could facilitate further research in this area, with neural networks playing a critical role in advancing our understanding of cancer genomics. AVAILABILITY AND IMPLEMENTATION: MUSE-XAE software is freely available at https://github.com/compbiomed-unito/MUSE-XAE.
Assuntos
Mutação , Neoplasias , Humanos , Neoplasias/genética , Algoritmos , Software , Genômica/métodos , Biologia Computacional/métodos , Redes Neurais de ComputaçãoRESUMO
Data sharing among different institutions represents one of the major challenges in developing distributed machine learning approaches, especially when data is sensitive, such as in medical applications. Federated learning is a possible solution, but requires fast communications and flawless security. Here, we propose SYNDSURV (SYNthetic Distributed SURVival), an alternative approach that simplifies the current state-of-the-art paradigm by allowing different centres to generate local simulated instances from real data and then gather them into a centralised hub, where an Artificial Intelligence (AI) model can learn in a standard way. The main advantage of this procedure is that it is model-agnostic, therefore prediction models can be directly applied in distributed applications without requiring particular adaptations as the current federated approaches do. To show the validity of our approach for medical applications, we tested it on a survival analysis task, offering a viable alternative to train AI models on distributed data. While federated learning has been mainly optimised for gradient-based approaches so far, our framework works with any predictive method, proving to be a comparable way of performing distributed learning without being too demanding towards each participating institute in terms of infrastructural requirements.
Assuntos
Inteligência Artificial , Aprendizado de Máquina , Análise de SobrevidaRESUMO
BACKGROUND: The impact of statin therapy on cardiovascular outcomes after ST-elevation acute myocardial infarction (STEMI) in real- world patients is understudied. AIMS: To identify predictors of low adherence and discontinuation to statin therapy within 6 months after STEMI and to estimate their impact on cardiovascular outcomes at one year follow-up. METHODS: We evaluated real-world adherence to statin therapy by comparing the number of bought tablets to the expected ones at 1 year follow-up through pharmacy registries. A total of 6043 STEMI patients admitted from 2012 to 2017 were enrolled in the FAST STEMI registry and followed up for 4,7 ± 1,6 years; 304 patients with intraprocedural and intrahospital deaths were excluded. The main outcomes evaluated were all-cause death, cardiovascular death, myocardial infarction, major and minor bleeding events, and ischemic stroke. The compliance cut-off chosen was 80% as mainly reported in literature. RESULTS: From a total of 5744 patients, 418 (7,2%) patients interrupted statin therapy within 6 months after STEMI, whereas 3337 (58,1%) presented >80% adherence to statin therapy. Statin optimal adherence (>80%) resulted as protective factor towards both cardiovascular (0.1% vs 4.6%; AdjHR 0.025, 95%CI 0.008-0.079, p < 0.001) and all-cause mortality (0.3% vs 13.4%; Adj HR 0.032, 95%CI 0.018-0.059, p < 0.001) at 1 year follow-up. Further, a significant reduction of ischemic stroke incidence (1% vs 2.5%, p = 0.001) was seen in the optimal adherent group. Statin discontinuation within 6 months after STEMI showed an increase of both cardiovascular (5% vs 1.7%; AdjHR 2.23; 95%CI 1.37-3.65; p = 0,001) and all-cause mortality (14.8% vs 5.1%, AdjHR 2.32; 95%CI 1.73-3.11; p ã0,001) at 1 year follow-up. After multivariate analysis age over 75 years old, known ischemic cardiopathy and female gender resulted as predictors of therapy discontinuation. Age over 75 years old, chronic kidney disease, previous atrial fibrillation, vasculopathy, known ischemic cardiopathy were found to be predictors of low statin adherence. CONCLUSIONS: n our real-world registry low statin adherence and discontinuation therapy within 6 months after STEMI were independently associated to an increase of cardiovascular and all-cause mortality at 1 year follow-up. Low statin adherence led to higher rates of ischemic stroke.
Assuntos
Inibidores de Hidroximetilglutaril-CoA Redutases , Adesão à Medicação , Sistema de Registros , Infarto do Miocárdio com Supradesnível do Segmento ST , Humanos , Infarto do Miocárdio com Supradesnível do Segmento ST/tratamento farmacológico , Infarto do Miocárdio com Supradesnível do Segmento ST/mortalidade , Masculino , Inibidores de Hidroximetilglutaril-CoA Redutases/administração & dosagem , Inibidores de Hidroximetilglutaril-CoA Redutases/uso terapêutico , Feminino , Adesão à Medicação/estatística & dados numéricos , Idoso , Pessoa de Meia-Idade , Seguimentos , Fatores de Tempo , Resultado do TratamentoRESUMO
Missense variation in genomes can affect protein structure stability and, in turn, the cell physiology behavior. Predicting the impact of those variations is relevant, and the best-performing computational tools exploit the protein structure information. However, most of the current protein sequence variants are unresolved, and comparative or ab initio tools can provide a structure. Here, we evaluate the impact of model structures, compared to experimental structures, on the predictors of protein stability changes upon single-point mutations, where no significant changes are expected between the original and the mutated structures. We show that there are substantial differences among the computational tools. Methods that rely on coarse-grained representation are less sensitive to the underlying protein structures. In contrast, tools that exploit more detailed molecular representations are sensible to structures generated from comparative modeling, even on single-residue substitutions.
Assuntos
Biologia Computacional , Mutação Puntual , Biologia Computacional/métodos , Proteínas/metabolismo , Estabilidade Proteica , Sequência de AminoácidosRESUMO
BACKGROUND: Central nervous system (CNS) infections in cattle are a major cause of economic loss and mortality. Machine learning (ML) techniques are gaining widespread application in solving predictive tasks in both human and veterinary medicine. OBJECTIVES: Our primary aim was to develop and compare ML models that could predict the likelihood of a CNS disorder of infectious or inflammatory origin in neurologically-impaired cattle. Our secondary aim was to create a user-friendly web application based on the ML model for the diagnosis of infection and inflammation of the CNS. ANIMALS: Ninety-eight cattle with CNS infection and 86 with CNS disorders of other origin. METHODS: Retrospective observational study. Six different ML methods (logistic regression [LR]; support vector machine [SVM]; random forest [RF]; multilayer perceptron [MLP]; K-nearest neighbors [KNN]; gradient boosting [GB]) were compared for their ability to predict whether an infectious or inflammatory disease was present based on demographics, neurological examination findings, and cerebrospinal fluid (CSF) analysis. RESULTS: All 6 methods had high prediction accuracy (≥80%). The accuracy of the LR model was significantly higher (0.843 ± 0.005; receiver operating characteristic [ROC] curve 0.907 ± 0.005 ) than the other models and was selected for implementation in a web application. CONCLUSION AND CLINICAL IMPORTANCE: Our findings support the use of ML algorithms as promising tools for veterinarians to improve diagnosis. The open-access web application may aid clinicians in achieving correct diagnosis of infectious and inflammatory neurological disorders in livestock, with the added benefit of promoting appropriate use of antimicrobials.
Assuntos
Doenças dos Bovinos , Doenças do Sistema Nervoso Central , Animais , Bovinos , Algoritmos , Doenças dos Bovinos/diagnóstico , Sistema Nervoso Central , Doenças do Sistema Nervoso Central/veterinária , Aprendizado de Máquina , Curva ROC , SoftwareRESUMO
Amyotrophic lateral sclerosis (ALS) is a highly complex and heterogeneous neurodegenerative disease that affects motor neurons. Since life expectancy is relatively low, it is essential to promptly understand the course of the disease to better target the patient's treatment. Predictive models for disease progression are thus of great interest. One of the most extensive and well-studied open-access data resources for ALS is the Pooled Resource Open-Access ALS Clinical Trials (PRO-ACT) repository. In 2015, the DREAM-Phil Bowen ALS Prediction Prize4Life Challenge was held on PRO-ACT data, where competitors were asked to develop machine learning algorithms to predict disease progression measured through the slope of the ALSFRS score between 3 and 12 months. However, although it has already been successfully applied in several studies on ALS patients, to the best of our knowledge deep learning approaches still remain unexplored on the ALSFRS slope prediction in PRO-ACT cohort. Here, we investigate how deep learning models perform in predicting ALS progression using the PRO-ACT data. We developed three models based on different architectures that showed comparable or better performance with respect to the state-of-the-art models, thus representing a valid alternative to predict ALS disease progression.
Assuntos
Esclerose Lateral Amiotrófica , Aprendizado Profundo , Doenças Neurodegenerativas , Esclerose Lateral Amiotrófica/diagnóstico , Esclerose Lateral Amiotrófica/terapia , Progressão da Doença , Humanos , Aprendizado de MáquinaRESUMO
The high cosine similarity between some single-base substitution mutational signatures and their characteristic flat profiles could suggest the presence of overfitting and mathematical artefacts. The newest version (v3.3) of the signature database available in the Catalogue Of Somatic Mutations In Cancer (COSMIC) provides a collection of 79 mutational signatures, which has more than doubled with respect to previous version (30 profiles available in COSMIC signatures v2), making more critical the associations between signatures and specific mutagenic processes. This study both provides a systematic assessment of the de novo extraction task through simulation scenarios based on the latest version of the COSMIC signatures and highlights, through a novel approach using archetypal analysis, which COSMIC signatures are redundant and more likely to be considered as mathematical artefacts. 29 archetypes were able to reconstruct the profile of all the COSMIC signatures with cosine similarity > 0.8. Interestingly, these archetypes tend to group similar original signatures sharing either the same aetiology or similar biological processes. We believe that these findings will be useful to encourage the development of new de novo extraction methods avoiding the redundancy of information among the signatures while preserving the biological interpretation.