RESUMEN
SUMMARY: Due to their flexibility and superior performance, machine learning models frequently complement and outperform traditional statistical survival models. However, their widespread adoption is hindered by a lack of user-friendly tools to explain their internal operations and prediction rationales. To tackle this issue, we introduce the survex R package, which provides a cohesive framework for explaining any survival model by applying explainable artificial intelligence techniques. The capabilities of the proposed software encompass understanding and diagnosing survival models, which can lead to their improvement. By revealing insights into the decision-making process, such as variable effects and importances, survex enables the assessment of model reliability and the detection of biases. Thus, transparency and responsibility may be promoted in sensitive areas, such as biomedical research and healthcare applications. AVAILABILITY AND IMPLEMENTATION: survex is available under the GPL3 public license at https://github.com/modeloriented/survex and on CRAN with documentation available at https://modeloriented.github.io/survex.
Asunto(s)
Inteligencia Artificial , Investigación Biomédica , Reproducibilidad de los Resultados , Programas Informáticos , Aprendizaje AutomáticoRESUMEN
The growing need for in-depth analysis of predictive models leads to a series of new methods for explaining their local and global properties. Which of these methods is the best? It turns out that this is an ill-posed question. One cannot sufficiently explain a black-box machine learning model using a single method that gives only one perspective. Isolated explanations are prone to misunderstanding, leading to wrong or simplistic reasoning. This problem is known as the Rashomon effect and refers to diverse, even contradictory, interpretations of the same phenomenon. Surprisingly, most methods developed for explainable and responsible machine learning focus on a single-aspect of the model behavior. In contrast, we showcase the problem of explainability as an interactive and sequential analysis of a model. This paper proposes how different Explanatory Model Analysis (EMA) methods complement each other and discusses why it is essential to juxtapose them. The introduced process of Interactive EMA (IEMA) derives from the algorithmic side of explainable machine learning and aims to embrace ideas developed in cognitive sciences. We formalize the grammar of IEMA to describe human-model interaction. It is implemented in a widely used human-centered open-source software framework that adopts interactivity, customizability and automation as its main traits. We conduct a user study to evaluate the usefulness of IEMA, which indicates that an interactive sequential analysis of a model may increase the accuracy and confidence of human decision making.
RESUMEN
Machine learning methods can detect complex relationships between variables, but usually do not exploit domain knowledge. This is a limitation because in many scientific disciplines, such as systems biology, domain knowledge is available in the form of graphs or networks, and its use can improve model performance. We need network-based algorithms that are versatile and applicable in many research areas. In this work, we demonstrate subnetwork detection based on multi-modal node features using a novel Greedy Decision Forest (GDF) with inherent interpretability. The latter will be a crucial factor to retain experts and gain their trust in such algorithms. To demonstrate a concrete application example, we focus on bioinformatics, systems biology and particularly biomedicine, but the presented methodology is applicable in many other domains as well. Systems biology is a good example of a field in which statistical data-driven machine learning enables the analysis of large amounts of multi-modal biomedical data. This is important to reach the future goal of precision medicine, where the complexity of patients is modeled on a system level to best tailor medical decisions, health practices and therapies to the individual patient. Our proposed explainable approach can help to uncover disease-causing network modules from multi-omics data to better understand complex diseases such as cancer.
Asunto(s)
Algoritmos , Aprendizaje Automático , Biología Computacional/métodos , Humanos , Medicina de Precisión , Biología de SistemasRESUMEN
PURPOSE: From April to September 2020, Poland was minimally affected by COVID-19 compared to other EU countries. We aimed to investigate the risks of false reverse transcription polymerase chain reaction (RT-PCR) results during the first wave (compared to later waves), that rises when cycle threshold (Ct) of positive result is close to limit of detection (LOD). MATERIALS/METHODS: We analyzed Ct values of SARS-CoV-2 positive RT-PCR results of 7726 patients in Poland from April-September 2020. SARS-CoV-2 positive RT-PCR results of 14,534 patients in the 2nd-3rd wave and 10,861 patients in the 4th-5th pandemic waves were used. Statistical analysis was based on one-way analysis of variance. To verify, 95% confidence intervals with Bonferroni correction were computed. Incidence of SARS-CoV-2 variants in Poland was analyzed using Whole Genome Sequencing from 923 (3.6%) patients. RESULTS: The mean Ct of RT-PCR positive test results analyzed ranged between 22.89 and 26.71 depending on the month of the results collection. The differences between months were significant (p â< â0.001). Differences in Ct were observed between age groups, with younger patients displaying higher Ct values, however, major trends over time were paralleled between age groups. CONCLUSIONS: The mean Ct of the tested RT-PCR positive test results was lower than 35 which is considered an upper borderline for reliable positive results of the assay. Therefore, most COVID-19 cases recorded in Poland from April to September 2020 were detected with minor risks of inaccuracy. Data from a single center exhibited greater consistency for both virus Ct level and SARS-CoV-2 virus variant identification.
Asunto(s)
COVID-19 , SARS-CoV-2 , Humanos , SARS-CoV-2/genética , Pandemias , COVID-19/diagnóstico , COVID-19/epidemiología , Reacción en Cadena de la Polimerasa de Transcriptasa Inversa , Polonia/epidemiología , Sensibilidad y EspecificidadRESUMEN
LncRNAs have arisen as new players in the world of non-coding RNA. Disrupted expression of these molecules can be tightly linked to the onset, promotion and progression of cancer. The present study estimated the usefulness of 14 lncRNAs (HAGLR, ADAMTS9-AS2, LINC00261, MCM3AP-AS1, TP53TG1, C14orf132, LINC00968, LINC00312, TP73-AS1, LOC344887, LINC00673, SOX2-OT, AFAP1-AS1, LOC730101) for early detection of non-small-cell lung cancer (NSCLC). The total RNA was isolated from paired fresh-frozen cancerous and noncancerous lung tissue from 92 NSCLC patients diagnosed with either adenocarcinoma (LUAD) or lung squamous cell carcinoma (LUSC). The expression level of lncRNAs was evaluated by a quantitative real-time PCR (qPCR). Based on Ct and delta Ct values, logistic regression and gradient boosting decision tree classifiers were built. The latter is a novel, advanced machine learning algorithm with great potential in medical science. The established predictive models showed that a set of 14 lncRNAs accurately discriminates cancerous from noncancerous lung tissues (AUC value of 0.98 ± 0.01) and NSCLC subtypes (AUC value of 0.84 ± 0.09), although the expression of a few molecules was statistically insignificant (SOX2-OT, AFAP1-AS1 and LOC730101 for tumor vs. normal tissue; and TP53TG1, C14orf132, LINC00968 and LOC730101 for LUAD vs. LUSC). However for subtypes discrimination, the simplified logistic regression model based on the four variables (delta Ct AFAP1-AS1, Ct SOX2-OT, Ct LINC00261, and delta Ct LINC00673) had even stronger diagnostic potential than the original one (AUC value of 0.88 ± 0.07). Our results demonstrate that the 14 lncRNA signature can be an auxiliary tool to endorse and complement the histological diagnosis of non-small-cell lung cancer.