Postmortem Interval Estimation Using Protein Chip Technology Combined with Multivariate Analysis Methods

Abstract

ObjectiveTo estimate postmortem interval （PMI） by analyzing the protein changes in skeletal muscle tissues with the protein chip technology combined with multivariate analysis methods.MethodsRats were sacrificed for cervical dislocation and placed at 16 ℃. Water-soluble proteins in skeletal muscles were extracted at 10 time points （0 d, 1 d, 2 d, 3 d, 4 d, 5 d, 6 d, 7 d, 8 d and 9 d） after death. Protein expression profile data with relative molecular mass of 14 000-230 000 were obtained. Principal component analysis （PCA） and orthogonal partial least squares （OPLS） were used for data analysis. Fisher discriminant model and back propagation （BP） neural network model were constructed to classify and preliminarily estimate the PMI. In addition, the protein expression profiles data of human skeletal muscles at different time points after death were collected, and the relationship between them and PMI was analyzed by heat map and cluster analysis.ResultsThe protein peak of rat skeletal muscle changed with PMI. The result of PCA combined with OPLS discriminant analysis showed statistical significance in groups with different time points （P<0.05） except 6 d, 7 d and 8 d after death. By Fisher discriminant analysis, the accuracy of internal cross-validation was 71.4% and the accuracy of external validation was 66.7%. The BP neural network model classification and preliminary estimation results showed the accuracy of internal cross-validation was 98.2%, and the accuracy of external validation was 95.8%. There was a significant difference in protein expression between 4 d and 25 h after death by the cluster analysis of the human skeletal muscle samples.ConclusionThe protein chip technology can quickly, accurately and repeatedly obtain water-soluble protein expression profiles in rats’ and human skeletal muscles with the relative molecular mass of 14 000-230 000 at different time points postmortem. The establishment of multiple PMI estimation models based on multivariate analysis can provide a new idea and method for PMI estimation.

Keywords

• forensic pathology
• postmortem interval
• protein chip
• protein expression profile
• fisher discriminant analysis
• back propagation neural network
• rats