Lang Dean H, Sharkey Neil A, Lionikas Arimantas, Mack Holly A, Larsson Lars, Vogler George P, Vandenbergh David J, Blizard David A, Stout Joseph T, Stitt Joseph P, McClearn Gerald E
Department of Kinesiology, College of Health and Human Development, The Pennsylvania State University, University Park, Pennsylvania 16802, USA.
J Bone Miner Res. 2005 May;20(5):748-57. doi: 10.1359/JBMR.041224. Epub 2004 Dec 20.
The aim of this study was to compare three methods of adjusting skeletal data for body size and examine their use in QTL analyses. It was found that dividing skeletal phenotypes by body mass index induced erroneous QTL results. The preferred method of body size adjustment was multiple regression.
Many skeletal studies have reported strong correlations between phenotypes for muscle, bone, and body size, and these correlations add to the difficulty in identifying genetic influence on skeletal traits that are not mediated through overall body size. Quantitative trait loci (QTL) identified for skeletal phenotypes often map to the same chromosome regions as QTLs for body size. The actions of a QTL identified as influencing BMD could therefore be mediated through the generalized actions of growth on body size or muscle mass.
Three methods of adjusting skeletal phenotypes to body size were performed on morphologic, structural, and compositional measurements of the femur and tibia in 200-day-old C57BL/6J x DBA/2 (BXD) second generation (F(2)) mice (n = 400). A common method of removing the size effect has been through the use of ratios. This technique and two alternative techniques using simple and multiple regression were performed on muscle and skeletal data before QTL analyses, and the differences in QTL results were examined.
The use of ratios to remove the size effect was shown to increase the size effect by inducing spurious correlations, thereby leading to inaccurate QTL results. Adjustments for body size using multiple regression eliminated these problems. Multiple regression should be used to remove the variance of co-factors related to skeletal phenotypes to allow for the study of genetic influence independent of correlated phenotypes. However, to better understand the genetic influence, adjusted and unadjusted skeletal QTL results should be compared. Additional insight can be gained by observing the difference in LOD score between the adjusted and nonadjusted phenotypes. Identifying QTLs that exert their effects on skeletal phenotypes through body size-related pathways as well as those having a more direct and independent influence on bone are equally important in deciphering the complex physiologic pathways responsible for the maintenance of bone health.
本研究的目的是比较三种针对体型调整骨骼数据的方法，并检验它们在数量性状基因座（QTL）分析中的应用。研究发现，用体重指数除骨骼表型会得出错误的QTL结果。体型调整的首选方法是多元回归。
许多骨骼研究报告了肌肉、骨骼和体型表型之间的强相关性，这些相关性增加了识别不受整体体型介导的对骨骼性状的遗传影响的难度。为骨骼表型鉴定的数量性状基因座（QTL）通常映射到与体型QTL相同的染色体区域。因此，被鉴定为影响骨密度（BMD）的QTL的作用可能是通过生长对体型或肌肉量的普遍作用来介导的。
对200日龄的C57BL/6J×DBA/2（BXD）第二代（F₂）小鼠（n = 400）的股骨和胫骨进行形态学、结构和成分测量，采用三种方法将骨骼表型调整为体型。一种常用的消除体型效应的方法是使用比率。在进行QTL分析之前，对肌肉和骨骼数据采用这种技术以及另外两种使用简单回归和多元回归的替代技术，并检验QTL结果的差异。
使用比率消除体型效应被证明会通过诱导虚假相关性而增加体型效应，从而导致不准确的QTL结果。使用多元回归进行体型调整消除了这些问题。应该使用多元回归来消除与骨骼表型相关的协变量的方差，以便研究独立于相关表型的遗传影响。然而，为了更好地理解遗传影响，应该比较调整后的和未调整的骨骼QTL结果。通过观察调整后的和未调整的表型之间的对数优势（LOD）得分差异可以获得更多见解。识别通过与体型相关的途径对骨骼表型产生影响的QTL以及那些对骨骼具有更直接和独立影响的QTL，对于解读负责维持骨骼健康的复杂生理途径同样重要。
