Bondi Danilo, Bevere Michele, Piccirillo Rosanna, Sorci Guglielmo, Di Felice Valentina, Re Cecconi Andrea David, D'Amico Daniela, Pietrangelo Tiziana, Fulle Stefania
Department of Neuroscience, Imaging and Clinical Sciences, University "G. d'Annunzio" Chieti - Pescara, Chieti, Italy; Interuniversity Institute of Myology (IIM), Perugia, Italy.
Department of Neuroscience, Imaging and Clinical Sciences, University "G. d'Annunzio" Chieti - Pescara, Chieti, Italy.
Mol Genet Metab. 2023 Nov;140(3):107705. doi: 10.1016/j.ymgme.2023.107705. Epub 2023 Oct 4.
Beyond classical procedures, bioinformatic-assisted approaches and computational biology offer unprecedented opportunities for scholars. However, these amazing possibilities still need epistemological criticism, as well as standardized procedures. Especially those topics with a huge body of data may benefit from data science (DS)-assisted methods. Therefore, the current study dealt with the combined expert-assisted and DS-assisted approaches to address the broad field of muscle secretome. We aimed to apply DS tools to fix the literature research, suggest investigation targets with a data-driven approach, predict possible scenarios, and define a workflow.
Recognized scholars with expertise on myokines were invited to provide a list of the most important myokines. GeneRecommender, GeneMANIA, HumanNet, and STRING were selected as DS tools. Networks were built on STRING and GeneMANIA. The outcomes of DS tools included the top 5 recommendations. Each expert-led discussion has been then integrated with an DS-led approach to provide further perspectives.
Among the results, 11 molecules had already been described as bona-fide myokines in literature, and 11 molecules were putative myokines. Most of the myokines and the putative myokines recommended by the DS tools were described as present in the cargo of extracellular vesicles.
Including both supervised and unsupervised learning methods, as well as encompassing algorithms focused on both protein interaction and gene represent a comprehensive approach to tackle complex biomedical topics. DS-assisted methods for reviewing existent evidence, recommending targets of interest, and predicting original scenarios are worth exploring as in silico recommendations to be integrated with experts' ideas for optimizing molecular studies.
除了传统方法外，生物信息学辅助方法和计算生物学为学者提供了前所未有的机遇。然而，这些惊人的可能性仍需要认识论批判以及标准化程序。特别是那些拥有大量数据的主题可能会受益于数据科学（DS）辅助方法。因此，本研究采用专家辅助和DS辅助相结合的方法来处理肌肉分泌组这一广泛领域。我们旨在应用DS工具来整理文献研究，以数据驱动的方法提出研究目标，预测可能的情况，并定义一个工作流程。
邀请了在肌动蛋白方面具有专业知识的知名学者提供最重要的肌动蛋白列表。选择了GeneRecommender、GeneMANIA、HumanNet和STRING作为DS工具。在STRING和GeneMANIA上构建网络。DS工具的结果包括前5条推荐。然后，将每次专家主导的讨论与DS主导的方法相结合，以提供更多视角。
在结果中，有11种分子在文献中已被描述为真正的肌动蛋白，11种分子为假定的肌动蛋白。DS工具推荐的大多数肌动蛋白和假定肌动蛋白被描述为存在于细胞外囊泡的货物中。
包括监督学习和无监督学习方法，以及涵盖专注于蛋白质相互作用和基因的算法，代表了一种处理复杂生物医学主题的综合方法。DS辅助方法用于审查现有证据、推荐感兴趣的目标和预测原始情况，作为与专家想法相结合以优化分子研究的计算机模拟推荐，值得探索。
