Endogenous and Exogenous Small RNA Signatures as Novel Tools for Postmortem Interval Determination

Figures

• 

  Figure 1. RIN Evaluation in Eight Mouse Organs and PMI Estimation within 48 Hours Post-Decapitation. (A) Average RIN values (n = 6) and standard deviations across eight organs at different postmortem time points. The line graph shows the trend of mean RIN values over time, with error bars indicating standard deviations at each time point. (B) Comparison of estimation performance for four regression models (LASSO, Elastic Net, RF, ANN) in predicting PMI (hours). Scatter plots show predicted vs. actual values, with the red diagonal line representing ideal estimation. Evaluation metrics include R2, MAE (hours) and RMSE (hours). (C) Importance ranking of organ features for PMI estimation across models. Normalized importance scores (sum = 1) highlight the heart (red) as the most important feature across all models, with other organs shown in blue. (D) SHAP value analysis of feature contributions for PMI estimation in LASSO, Elastic Net, RF, and ANN models. The heart feature is shown in red, with standardized feature values visualized as “High-Low”.

• 

  Figure 2. Dynamic Small RNA Profiling in Cardiac Tissue within 30 Hours Postmortem by PANDORA-seq. (A) Proportion of sRNAs mapped to the mouse reference genome (average, n = 5). (B) Distribution of sRNA fragment lengths mapped to the mouse reference genome (average, n = 5). (C) Proportion of mRNA fragments mapped to the mouse mRNA reference sequence (average, n = 5). (D) Distribution of mRNA fragment lengths mapped to the mouse mRNA reference sequence (average, n = 5). (E) Length distribution of five sRNA types at six time points (average, n = 5). (F) Expression trends of various sRNAs over time (RPM, Reads Per Million), normalized. * p < 0.05; ** p < 0.01, comparing data to 0 h (t-test with Benjamini–Hochberg FDR correction). The x-axis represents time (hours) and the y-axis shows RPM (mean ± SD, n = 5), with shaded areas indicating standard deviation.

• 

  Figure 3. Dynamic Analysis of miRNA Expression in Postmortem Mouse Cardiac Tissue. (A) Fuzzy membership values for miRNAs assigned to six clusters (Cluster 1–6) at six postmortem time points (0 h, 6 h, 12 h, 18 h, 24 h, 30 h). Rows represent individual miRNAs, and columns represent PMI time points. Color intensity indicates the degree of membership (scale: 0 to 1). Each cluster peaks at different PMI time points. (B) Relative abundance of miRNAs within each cluster as a percentage of the total miRNAs across all clusters. (C–G) Gene Ontology (GO) enrichment analysis of target genes for miRNAs in (C) Cluster 2 (peak at 6 h), (D) Cluster 3 (peak at 12 h), (E) Cluster 4 (peak at 18 h), (F) Cluster 5 (high at 24–30 h), and (G) Cluster 6 (peak at 30 h). Each plot shows the enriched biological processes for the corresponding miRNA cluster.

• 

  Figure 4. Biomarker Screening and Model Construction for PMI Estimation Based on PANDORA-seq, PCR Array, and Machine Learning. (A) Workflow for selecting biomarkers to efficiently predict PMI. Red boxes represent candidate sncRNAs or biomarkers, and blue boxes represent screening technologies. (B) PCR Array results for 20 candidate sncRNAs. The x-axis represents PMI (in minutes), and color intensity indicates expression abundance (Z-score normalized). (C) Comparison of performance across nine machine learning models using PCR Array data from 20 sncRNAs. The three axes represent the model’s R2, MAE and RMSE (MAE and RMSE in minutes). (D) Final selection of seven key biomarkers (4 tsRNAs and 3 piRNAs, marked in red) identified using recursive feature elimination within a nested cross-validation framework and 15-fold cross-validation. (E) Performance evaluation of the final RF model constructed using the seven biomarkers, with the red diagonal line representing ideal predictions. The upper left corner shows evaluation metrics, with MAE and RMSE in minutes. (F) Importance score ranking of the seven biomarkers in the final RF model. (G) Feature importance ranking (Mean Absolute SHAP Value) for…

• 

  Figure 5. Expression Characteristics of Microbial-derived sRNAs at Postmortem Time Points. (A) Heatmap showing read counts for 41 microbial-specific differential fragments, with red indicating high expression and blue indicating low expression. Data were Z-score normalized to eliminate sequence-specific differences. (B) Species composition of the 41 microbial-specific differential fragments. (C) Line plot illustrating changes in the matching rate between sRNAs and the Staphylococcus aureus genome from 0 to 30 h postmortem (mean ± SD, n = 5). X-axis: postmortem time (h); Y-axis: matching rate (%). The blue line denotes the mean and the light-blue shading represents the standard deviation range.

• 

  Figure 6. Endogenous Cardiac miRNAs Targeting Staphylococcus aureus mRNAs Postmortem. (A) Bar plots showing the number of predicted Staphylococcus aureus mRNA targets for highly expressed miRNAs in Clusters 1–5. In each subplot, the miRNA with the highest target count is highlighted in red, while the others are shown in blue. (B) Heatmap illustrating the predicted binding of miRNAs to S. aureus mRNA targets. Color intensity reflects the proportion of target counts, with darker shades indicating higher values. Each cell is annotated with its corresponding percentage.