基于高通量mRNA差异表达分析构建SEM-GC模型推断损伤时间的研究

Wound age estimation is one of the most critical and difficulty issues for the forensic pathologist in a violent death, but so far there isn’t an accurately and stable method to estimate the time interval after injury, because there are many factors influenced the tissues repairs after injury. We have performed a study on mRNA difference expression after skeletal muscle contused in rats by Digital Gene Expression. The results show that there was a large amount of difference expression mRNAs discovered after skeletal muscle contused. It suggests that these mRNA could act as the biomarker to estimate the time interval after injury. But the following problems should be solved before implementing these indicators to estimate wound age. First, there are too many mRNA difference genes should to be detected, which have complex correlation relationships with each other. It is very necessary to classify them and find some representatives indicators by using the bioinformatics analysis. Second, these data of genes with difference expression has shown complicated liner or non-liner time trend, as well as multilevel data character. So a synthetic analysis method with the capability of dealing with multilevel and longitudinal continuous data should be considered. In present study, we plan to establish the animal model of skeletal muscle contused in rats, with considering the influence factors, such as gender, age, degree of injury, time interval after injury, postmortem interval and the environment after death, etc. The representative’s indicators which have a high correlation relationship with the time interval will be filtered by high-through Real-time PCR, logic determines, GO analysis and Pathway analysis. Finally, we plan to implement the Structure Equation Model Growth Curve (SEM-GC) method to fit a mathematical model for analyzing the data synthetically. As a result, we hope it is a novel analysis strategy for the wound age estimation.

在暴力致死、多次损伤致死案件中，由于影响因素众多很难做到损伤时间准确推断，影响案件侦破和审理。我们通过研究发现肌肉损伤后有大量mRNA出现差异表达，且呈现一定程度的时序性变化，可以用于损伤时间推断。但仍存在以下问题①指标繁多，各指标之间存在复杂相关性，需要进一步归类分析；②各差异表达基因在时序变化上表现为复杂的线性/非线性时序关系，且在时序变化方面具有多水平资料的特征,需要一种能处理多水平复杂时序变化资料的分析策略。因此，本课题拟建立不同年龄、不同损伤程度、不同损伤时间、不同死后保存时间及保存条件的大样本动物损伤模型，通过检测识别、逻辑判断、生物信息学功能分析、因子分析法筛选与损伤时间高度相关的指标，运用高通量Real-time PCR技术进行定量分析，建立运用mRNA差异表达分析构建SEM-GC模型预测个体损伤时间的研究方法，为后期进行人类损伤时间推断提供新的分析思路。

损伤时间推断是一个复杂的多因素问题，目前仍没有一个可靠、可行的方法对的损伤时间进行较为准确的推断。该课题建立了不同年龄、不同损伤程度、不同损伤时间、不同死后保存时间及保存条件的动物损伤模型，采用RT-qPCR对筛选的 mRNA进行了检测。我们尝试利用多个指标表达量数据结合多元统计技术建立了余弦相似性、Fisher判别及 偏最小二乘（PLS）回归模型，对损伤时间进行预测，结果显示，模型具有较为稳定的可靠性，可对损伤时间进行较为准确的推断，认为数学模型的方法可应用于损伤时间推断是可行的。获得的数据结果显示，同一指标损伤后变化在不同季节随死亡时间变化是不同的，不同指标损伤后变化在同一季节随死亡时间变化也是不同的，不同年龄指标的随损伤时间表达变化也是不同的，因此需要对多个指标进行综合分析，前期探索的利用多元的深度学习算法建立数学模型的方法能够为多因素影响下损伤时间推断提供新的思路。