警惕所谓的“良好”相关性:对个体 mRNA-蛋白质预测的统计现实检验
Beware of so-called 'good' correlations: a statistical reality check on individual mRNA-protein predictions
AI总结:
本文从统计学角度警示,仅凭 mRNA-蛋白质正相关不足以用单一 mRNA 定量推断对应蛋白质的丰度或功能,并揭示了给定相关系数下蛋白质水平的巨大变异性及分布两端错配现象。
摘要:
生命科学研究常将信使核糖核酸(mRNA)定量作为功能分析的独立方法。然而,尽管实测的 mRNA 与蛋白质水平之间呈正相关,经验观察到的相关系数并不完整,因此在做出不可知论推断时须保持谨慎。本文对转录组学-蛋白质组学研究中相关性强度的概念进行了统计反思并提出警示。文章强调,在给定的经验相关值下,即使 mRNA 含量精确,可能的蛋白质水平仍存在较大变异性,并指出有相当比例的 mRNA-蛋白质对的丰度位于各自分布的两端。细胞生物学家、数据科学家和生物统计学家应认识到,仅凭 mRNA-蛋白质相关性不足以支持使用单一 mRNA 定量来推断其对应蛋白质的数量或功能。
英文摘要:
Research in the life sciences often employs messenger ribonucleic acids (mRNA) quantification as a standalone approach for functional analysis. However, although the correlation between the measured levels of mRNA and proteins is positive, correlation coefficients observed empirically are incomplete, necessitating caution in making agnostic inferences. This essay provides a statistical reflection and caveat on the concept of correlation strength in the context of transcriptomics-proteomics studies. It highlights the variability in possible protein levels at given empirical correlation values, even for precise mRNA amount, and underscores the notable proportion of mRNA-protein pairs with abundances at opposite ends of their respective distributions. Cell biologists, data scientists, and biostatisticians should recognise that mRNA-protein correlation alone is insufficient to justify using a single mRNA quantification to infer the amount or function of its corresponding protein.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。