Constructing a cross-species injury time inference model and injury time point prediction method

Abstract

This invention belongs to the field of forensic injury pathology, specifically relating to the construction of cross-species injury time inference models and injury time point prediction methods. Utilizing advanced bioinformatics techniques, starting from orthologous or paralogous genes, and based on transcriptomics, it applies time-conserved cross-species molecules to cross-species injury time inference, improving the accuracy and reliability of predicting skeletal muscle injury recovery processes. This invention integrates multiple analytical tools and algorithms, enabling comprehensive exploration of cross-species conserved temporal patterns and providing a more comprehensive analytical perspective. It reduces cross-species bias effects: through methods such as Combat, it effectively reduces the cross-species bias effect of biological data, ensuring the accuracy of the results. This invention deeply explores cross-species conserved temporal patterns, reduces the cross-species bias effect of biological data, deeply explores the conservation of biomarkers, and establishes a comprehensive time-series cross-species research framework.

Description

Technical Field

[0001] This invention belongs to the field of forensic injury pathology, and in particular relates to the construction of a cross-species injury time inference model and a method for predicting injury time points. Background Technology

[0002] With the increasing prevalence of traffic accidents, violent conflicts, accidental injuries, and sports-related incidents, skeletal muscle injuries are becoming more common. Severe injuries not only significantly impact the quality of life of injured individuals but also prolong recovery time. Therefore, accurately estimating the time of injury is crucial not only for patient treatment but also for forensic medicine, sports medicine, and public safety. However, due to ethical constraints, we cannot directly study the repair process of human skeletal muscle injuries. Traditional research methods have low throughput and are concentrated on single model organisms, making it difficult to reveal complex gene regulatory mechanisms. Most studies on skeletal muscle injuries can only be conducted on animal models and cannot be directly applied to humans. Therefore, how to transfer the large amount of experimental data from animal models to humans and achieve accurate prediction of human biological information is a current major challenge in forensic research.

[0003] Existing cross-species studies are mostly limited to comparisons within a single species, often focusing only on biomarkers at a single time point and lacking in-depth analysis of co-expressed genes. This limits a comprehensive understanding of the injury recovery process and makes it difficult to effectively translate data from different species across species. Furthermore, existing cross-species comparison methods often ignore biological differences between species, resulting in insufficient comparability of research results.

[0004] Therefore, it is particularly important to break down the biological barriers between species and fill this gap through systematic cross-species research. Summary of the Invention

[0005] To address the current inability to accurately predict the timing of skeletal muscle injuries in humans, this invention provides a method for constructing a cross-species injury timing estimation model and predicting injury time points. Leveraging the advantages of transcriptomics, this invention starts with cross-species homologous genes, deeply explores temporally conserved biological indicators across species, and employs methods such as ComBat to reduce the cross-species bias effect in biological data. This establishes a comprehensive temporal-based cross-species estimation model to improve the understanding and prediction of the skeletal muscle injury recovery process. This innovative approach will not only provide scientific and objective methodological support for cross-species injury timing estimation but will also promote the development of biomedicine and forensic medicine, filling gaps in current research. With the increasing urgency of the need for injury timing estimation, this invention will provide important technical support and theoretical basis for related fields.

[0006] This invention is achieved through the following technical solution: a method for establishing a cross-species damage time inference model, comprising the following steps:

[0007] ① First, construct at least two different animal models of skeletal muscle contusion at different injury times and collect tissue samples; extract total RNA from these tissues, sequence the samples, and finally generate transcriptome expression profiles of skeletal muscle injury time points in different animals.

[0008] ②Analyze and screen orthologous or paralogous genes of different animals and humans to obtain conserved homologous genes across species;

[0009] ③We performed weighted gene co-expression network analysis on the transcriptome expression profiles of different animals at different time points of skeletal muscle injury within their respective species; we merged the inter-species genes at different time points to obtain a set of co-expressed genes at different injury time points across species.

[0010] ④ By utilizing the set of co-expressed genes at different injury time points across species, we can obtain protein interaction networks from different animals and humans, perform deep chimerism of cross-species protein expression networks, and further obtain conserved biomarkers related to cross-species skeletal muscle injury repair.

[0011] ⑤ Introduce the human skeletal muscle injury dataset from a public database as an external validation of the cross-species model for skeletal muscle injury time inference;

[0012] ⑤ Using the expression profiles of the conserved biological indicators related to skeletal muscle injury repair constructed above, a time-series conserved biological indicator machine learning model was constructed.

[0013] As a further improvement to the technical solution of the method established in this invention, before step ⑤, conserved biological indicators related to skeletal muscle injury repair across species are merged to reduce the cross-species bias of biological data.

[0014] As a further improvement to the technical solution of the method of the present invention, in step ④, the Gosline network alignment tool is used to obtain protein interaction networks of different animals and humans.

[0015] As a further improvement to the technical solution of the method established in this invention, when reducing cross-species offset of biological data, the Combat, Limma, Batch Mean centering or Genenorm methods are used.

[0016] As a further improvement to the technical solution of the method established in this invention, in step ②, when analyzing and screening different animals and humans, a pairwise gene comparison screening method is used to obtain the intersection and acquire cross-species conserved homologous genes. When multiple species are compared and screened pairwise for orthologous genes, such as mouse-pig orthologous gene screening, mouse-human orthologous gene screening, and pig-human orthologous gene screening, and the intersection is obtained, the common orthologous genes of the studied multiple species are acquired.

[0017] As a further improvement to the technical solution of the method of the present invention, in step ③, when performing weighted gene co-expression network analysis, the module with the strongest positive correlation with each injury time point is selected within each species, and then the intersection of the gene sets of the corresponding time point modules across species is taken to obtain the gene set of co-expression at the injury time point across species.

[0018] As a further improvement to the technical solution of the method of the present invention, the machine learning model includes random forest, support vector machine, multilayer perceptron, and domain-adaptive deep learning algorithm.

[0019] As a further improvement to the technical solution of the method of the present invention, the order of steps ②, ③ and ④ can be replaced by ④, ②, ③, or ③, ④, ②.

[0020] It should be noted that when the order is ④, ②, ③, in step ④, when obtaining the protein interaction networks of different animals and humans, the genes of the animal proteins targeted are the transcriptome expression profiles of different animal skeletal muscle injury time points in step ①, to obtain another set of conserved biomarkers related to cross-species skeletal muscle injury repair; then steps ② and ③ are performed.

[0021] As a further improvement to the technical solution of the method of this invention, in step ⑤, a public database of human skeletal muscle injury dataset is introduced as an external validation of the cross-species model for inferring skeletal muscle injury time.

[0022] This invention also provides a method for predicting the time point of cross-species damage, which utilizes a machine learning model obtained by the aforementioned method for constructing a cross-species damage time inference model. The specific steps include:

[0023] By inputting conserved bioindicators related to skeletal muscle injury repair from human skeletal muscle injury samples across species into a machine learning model, cross-species predictions of the time points of human skeletal muscle injury can be obtained.

[0024] Compared with existing technologies, the method for constructing a cross-species damage time inference model and predicting damage time points proposed in this invention has the following advantages:

[0025] 1) Improved scientific rigor and objectivity: By using advanced bioinformatics technology, starting from orthologous or paralogous genes, and based on the transcriptome, temporally conserved molecules across species are applied to the inference of cross-species injury time, which improves the accuracy and reliability of the prediction of skeletal muscle injury recovery process.

[0026] 2) Integration of Multiple Analytical Tools: The method described in this invention integrates multiple analytical tools and algorithms, enabling comprehensive exploration of conserved temporal patterns across species and providing a more comprehensive analytical perspective. Reduction of Cross-Species Bias Effect: Through methods such as Combat, studies have effectively reduced the bias effect of biological data in cross-species analysis, ensuring the accuracy of the results.

[0027] 3) Establish a comprehensive research framework: By integrating various analytical tools and algorithms through the method described in this invention, starting from orthologous or paralogous genes, we can deeply explore the conserved temporal patterns across species, reduce the cross-species bias effect of biological data, deeply explore the conservation of biomarkers, and establish a comprehensive temporal cross-species research framework. Attached Figure Description

[0028] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 The implementation process of the method for constructing a cross-species damage time inference model involved in this invention is shown.

[0031] Figure 2 This illustrates the correlation between cross-species co-expression modules and sample features. Darker colors indicate stronger correlations. Within each cell, the top number represents the correlation coefficient (R-value), while the bottom numbers indicate the significance of the correlation (P-value).

[0032] Figure 3 Deep chimerism of rat-pig-human cross-species protein expression networks was performed using the Gosline network alignment tool. Connections represent cross-species proteins with similar expression levels.

[0033] Figure 4 Principal component analysis (PCA) plots of sample distributions for rats, pigs, and humans before and after Combat treatment.

[0034] Figure 5 The results represent the external validation of the cross-species model. The confusion matrix demonstrates the accuracy of injury time determination, while the ROC curve evaluates the model's performance. Detailed Implementation

[0035] To better understand the above-mentioned objectives, features, and advantages of the present invention, the solutions of the present invention will be further described below. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0036] Many specific details are set forth in the following description in order to provide a full understanding of the invention, but the invention may also be practiced in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of the invention, and not all embodiments.

[0037] The specific embodiments of the present invention will be described in detail below.

[0038] 1. Grouping of experimental animals

[0039] This study used 28 male Sprague Dawley (SD) rats aged 6 to 7 weeks (average weight 240 g ± 20 g), all provided by the Experimental Animal Center of Shanxi Medical University. These rats were randomly divided into a control group and an injury group. The control group consisted of uninjured skeletal muscle rats, while the injury group included rats divided into groups at 4 hours, 24 hours, and 48 hours after injury, with 7 rats in each group.

[0040] This invention selected four 9-month-old Bama pigs (weighing 20.06±1.14 kg), all provided by Beijing Shichuang Century Small Pig Breeding Base. The Bama pigs were randomly divided into a control group and an injury group. The control group consisted of Bama pigs with no skeletal muscle damage, while the injury group included groups at 4h, 24h, and 48h after injury, with one pig in each group.

[0041] 2. Preparation of animal models of skeletal muscle injury

[0042] After a 12-hour fast, experimental rats were anesthetized by intraperitoneal injection of 3% sodium pentobarbital (0.13 ml / 100 g). A rat skeletal muscle contusion model (injury group) was established by dropping a 500 g weight from a height of 50 cm onto the right hind limb. The rats were then fed and watered regularly. At 4 h, 24 h, and 48 h post-injury, the animals were euthanized with a lethal dose of 10% chloral hydrate (4 mL / kg body weight). The control group received no treatment and was treated in the same manner. Subsequently, approximately 100 mg of muscle tissue was harvested from the wound site of each rat. After removing the fascia, the samples were immediately frozen in liquid nitrogen and stored at -80 °C.

[0043] After fasting for 12 hours, Bama pigs were restrained and injected intramuscularly with atropine (1 ml, 0.5 mg) at a dose of 0.1 mg / kg, and the animals' reactions were observed. When the animals were in a relaxed state, they were injected intramuscularly with Sutamethasone 50 (a mixture of pentobarbital and chlorpromazine) (50 mg / ml lyophilized powder) at a dose of 5 mg / kg. The animals' respiration, physiological state, and body temperature changes were closely monitored, and they were kept warm. After the animals were sedated, their fore and hind limbs were secured with ropes, their mouths were secured with restraints, and their abdomens were elevated with cushioning pads and multi-layered foam pads to fully avoid injury to the abdomen and internal organs during impact. The pigs were subjected to a free fall of a 2 kg weight from a height of 1.66 m, which resulted in four injuries on each side of the spine to create a skeletal muscle injury model. The pigs were then euthanized, and tissue samples were directly taken from the corresponding sites in the injury group. Animals in the control group did not suffer skeletal muscle injuries and were euthanized and treated in the same way. Approximately 100 mg of muscle tissue samples were excised from the wounds of each Bama pig. After the fascia was removed, the sample was immediately frozen in liquid nitrogen and stored at -80 °C.

[0044] 3. Total RNA extraction and quality control of samples

[0045] Following the manufacturer's instructions, approximately 50 mg of muscle tissue sample was placed in an RNase-free mortar and ground in liquid nitrogen at a low temperature. Lysis was then performed using RNAiso Plus 9108 (Takara Bio, Shiga, Japan). The total RNA concentration (ng / ml) and purity were determined using a microplate reader (Infinite M200 Pro; TECAN, Zurich, Switzerland), and RNA quality was initially checked using a Nanodrop 2000 spectrophotometer (Thermo, USA). Samples with RNA absorbance (OD260 / 280) between 1.8 and 2.2 were suitable for subsequent experiments. The integrity of the total RNA was assessed using an Agilent 2100 Nano 6000 assay kit (Agilent Technologies, Palo Alto, CA, USA) via an Agilent 2100 Bioanalyzer. All samples showed an RNA integrity value (RIN) greater than 7.0, indicating sufficient quality for poly-A selection and cDNA library preparation.

[0046] 4. Transcriptome library construction, next-generation sequencing, and preliminary data analysis

[0047] Quality-controlled RNA samples were sent to the work platform for transcriptome library construction using the NEBNextUltra™ RNA Library Preparation Kit (polyA Select, NEB, Beijing, China). Preliminary quantification of the library was performed using a Qubit 3.0 Fluorometer (Life Technologies). Insert sizes in the transcriptome library were measured using the Agilent 2100 RNA6000 analysis kit, and accurate quantification of the constructed library was performed using the Bio-RAD KitIQ SYBR GRN (Bio-Rad Laboratories Inc., Hercules, CA, USA).

[0048] Following the mRNA high-throughput sequencing library preparation protocol, reverse transcription of RNA is first performed to synthesize the first and second strands. Then, adapters are added to the 3' end, and the 5' end is padded. After bridging PCR amplification, further purification, screening, and sequence quality assessment are required to improve the accuracy of the library.

[0049] We performed paired-end sequencing on the HiSeq-2000 platform (Illumina Inc., San Diego, CA, USA) using 150 bp sequences with a target coverage of 40-60M and a sequencing length of PE150. The muscle tissue sample library was sequenced using the HiSeq PE Cluster Kitv4-cBot hS (Illumina, USA). After sequencing, base identification was performed using bcl2fastq2 software to convert the offline data into raw sequence reads, and the results were saved in FASTQ file format.

[0050] Next, a series of bioinformatics analyses were performed on the open-source platform Chipster (version 3.16). FastQC software was used to comprehensively assess the quality of the sequencing data, and HISAT2 (version 2.1.0) was selected to map the processed reads onto the genome. Annotation was performed based on the files Rattus_norvegicus.Rnor_6.0.954 and Sus_scrofa.Sscrofa11.1.95. HTSeq (version 0.6.1) was used to quantify and match the reads, using default settings and unique matching for the reverse strand to obtain gene-level counts, resulting in the original gene expression matrices for each time point after injury and for the normal group. Finally, the original expression matrices were processed using the R Bioconductor package DESeq2 and normalized using DESeq. Genes with a total count below 100 in all samples were considered low-expression genes and removed.

[0051] 5. Bioinformatics Analysis

[0052] (1) Comparison of homologous genes between mice, pigs, and humans

[0053] After the above processing, the original expression profile was obtained. Then, mouse-pig-human homologous gene alignment was performed. The g:Orth tool was used to retrieve information from the Ensembl database to provide homologous genes, which were then used to identify homologous genes among rats, pigs, and humans in all three species. A total of 6367 orthologous genes from the three species were obtained. Subsequently, weighted gene co-expression network analysis was performed.

[0054] (2) Weighted Gene Co-expression Network Analysis (WGCNA)

[0055] We performed a weighted gene co-expression network analysis (WGCNA) on 6367 homologous genes from rats and pigs. Using the WGCNA package in R Studio, we identified gene co-expression modules at each post-injury time point. A soft threshold of β=22 was used for rats to obtain scale-free topology, while a soft threshold of β=12 was used for pigs. Topological overlap was calculated using an adjacency matrix to measure network interconnectivity. Applying WGCNA's dynamic tree-cutting algorithm, rats identified 7 modules, including 6 meaningful modules and one undefined gray module; pigs identified 16 modules, including 15 meaningful modules and one undefined gray module. Subsequently, Spearman correlation analysis was used to calculate the R-value and p-value of each module, and the module with the strongest positive correlation at each time point (i.e., the largest R-value) was selected and termed the High Correlated Module (HCr) at each time point. Based on the WGCNA analysis results, we ultimately selected 11 co-expression modules. Figure 2 Different colors are used to represent them.

[0056] The genes corresponding to the same injury time point in rats and pigs were intersected, and then all intersecting genes were merged. Among them, the normal group (C) after rat and pig skeletal muscle injury had 19 cross-species conserved genes, the 4h group had 140 genes, the 24h group had 104 genes, and the 48h group had 22 genes, resulting in a total of 285 cross-species conserved co-expressed gene indicators.

[0057] 6. Further extraction of conserved biomarkers related to cross-species skeletal muscle injury repair

[0058] In the aforementioned study, we identified 285 genes co-expressed and time-dependently expressed in the early stages of skeletal muscle injury in rats and pigs. These genes were conserved across different related species. Furthermore, to investigate the interspecies conservation of these genes at the protein expression level, we obtained protein interaction networks from rats, Bama pigs, and humans. Using the PPI network local alignment analysis (GASOLINE) algorithm, we achieved deep chimerism of the cross-species protein expression networks, exploring potentially conserved expressed proteins in rats, pigs, and humans, based on default parameters such as... Figure 2 Analysis revealed 20 conserved protein-protein interaction networks after skeletal muscle injury in rat, pig, and human species. Figure 3 Only the first two conserved networks out of the 20 networks are shown. After alignment with the protein network, the conservation of cross-species biological indicators is further enhanced. At this point, 105 conserved biological indicators related to skeletal muscle injury repair in rats, pigs, and humans are obtained.

[0059] 7. Include human skeletal muscle injury transcriptome data

[0060] We found the human skeletal muscle injury datasets GSE45426 and GSE107934 in the GEO database, and selected 49 samples from them, including the normal control group, the 4 h post-injury group, and the 24 h post-injury group, as external validations for the cross-species model for skeletal muscle injury time inference.

[0061] 8. Reduce interspecies effects

[0062] Due to differences in gene expression levels among species, we used four methods (Combat, Limma, BatchMean centering, and Genenorm) to reduce interspecies effects and allow for the analysis of interspecies data on the same dimension. Principal component analysis (PCA) was used to illustrate the data distribution at different time points after skeletal muscle injury in rats and pigs. Figure 3Principal component analysis results of biological data before and after cross-species expression shift reduction based on the Combat method are presented. After reducing interspecies effects, samples at different time points after skeletal muscle injury show temporal clustering, and the cross-species shift effect of transcriptome data is greatly reduced.

[0063] 9. Machine Learning Models

[0064] Using 105 conserved biomarkers related to skeletal muscle injury repair across species, we were able to infer the time point of skeletal muscle injury across species.

[0065] (1) Machine learning model

[0066] Using the expression profiles of conserved biomarkers related to skeletal muscle injury repair constructed above, we built a random forest (RF) cross-species machine learning model on the Anaconda (version 3) platform in Python. During internal validation, five-fold cross-validation was used to evaluate the model's accuracy (Acc), precision (Pre), recall (Rec), and F1 score. The sample data was randomly divided into a 70% training set and a 30% validation set, and the optimal parameters were determined using a grid search method (i.e., a strategy to find the optimal parameters among each machine learning algorithm). To further test the reliability of our cross-species research strategy, we modeled the model using skeletal muscle injury samples from rats and pigs, and performed external validation using human samples, achieving cross-species injury time-point prediction.

[0067] Without Combat mitigation of interspecies effects, the model combining 105 temporal cross-species conserved biological indicators with a random forest achieved an accuracy of 0.917 and an AUC of 0.991 in internal validation. However, in external validation, the accuracy was 0.551 and the AUC was 0.494. The confusion matrix showed that both 4-hour and 24-hour samples in the external validation were incorrectly classified as the normal group C. Figure 5 It is still not feasible to rely solely on conserved biomarkers across species for cross-species inference, and it is necessary to reduce cross-species expression bias in biological data.

[0068] Subsequently, after Combat reduction of interspecies effects, we found that the accuracy and AUC of 105 conserved biomarkers related to skeletal muscle injury repair across species, combined with the random forest model, were both 1.000 in internal validation and 1.000 in external validation.

[0069] according to Figure 4As shown in the multi-species PCA plot, after Combat reduction of inter-species effects, the biological data shifted from species clustering to time-point clustering, significantly reducing inter-species effects. This resulted in 100% accuracy in cross-species machine learning for determining time points. This further illustrates that our cross-species information transfer strategy framework can achieve cross-species prediction of damage time points.

[0070] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the present invention. Although detailed descriptions have been provided with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments, and they should all be covered within the protection scope of the claims.

Claims

1. A method for constructing a cross-species damage time inference model, characterized in that, Includes the following steps: ① First, construct at least two different animal models of skeletal muscle contusion at different injury times and collect tissue samples; extract total RNA from these tissues, sequence the samples, and finally generate transcriptome expression profiles of skeletal muscle injury time points in different animals. ②Based on the transcriptome expression profiles of skeletal muscle injury time points in different animals, we conducted orthologous or paralogous gene analysis and screening for different animals and humans to obtain conserved homologous genes across species. ③Weighed gene co-expression network analysis was performed on homologous genes conserved across species within their respective species; inter-species genes at different time points were merged to obtain a set of co-expressed genes at cross-species injury time points. ④ By utilizing the set of co-expressed genes at different injury time points across species, we can obtain protein interaction networks in different animals and humans, and perform deep chimerism of cross-species protein expression networks to further obtain conserved biomarkers related to cross-species skeletal muscle injury repair. ⑤ Using the expression profiles of the conserved biological indicators related to skeletal muscle injury repair constructed above, a time-series conserved biological indicator machine learning model was constructed. Before step ⑤, conserved bioindicators related to skeletal muscle injury repair across species are merged to reduce cross-species bias in the biological data. When reducing cross-species bias in the biological data, the Combat, Limma, Batch Mean centering or Genenorm methods are used.

2. The method for establishing a cross-species damage time inference model according to claim 1, characterized in that, In step ④, protein interaction networks from different animals and humans are obtained from the String database, and deep chimerism of cross-species protein expression networks is achieved using the Gasoline network alignment tool.

3. The method for establishing a cross-species damage time inference model according to claim 1, characterized in that, In step ②, when analyzing and screening different animals and humans, the g:Profiler R package is used to screen homologous genes in pairs of species, and finally a set of cross-species conserved homologous genes from multiple animal and human species is obtained.

4. The method for establishing a cross-species damage time inference model according to claim 1, characterized in that, In step ③, when performing weighted gene co-expression network analysis, the module with the strongest positive correlation to each injury time point is selected within each species, and then the intersection of the gene sets of the corresponding time points of the modules across species is taken to obtain the cross-species injury time point co-expression gene set.

5. The method for constructing a cross-species damage time inference model according to claim 1, characterized in that, The machine learning models include random forest, support vector machine, multilayer perceptron, and domain-adaptive deep learning algorithm.

6. The method for establishing a cross-species damage time inference model according to claim 1, characterized in that, In step ⑤, a public database of human skeletal muscle injury dataset is introduced as an external validation of the cross-species model for skeletal muscle injury time inference.

7. A method for predicting the time point of cross-species damage, characterized in that, The machine learning model used is obtained by the method for constructing a cross-species damage time inference model as described in any one of claims 1-6, and the specific steps include: By inputting conserved bioindicators related to skeletal muscle injury repair from human skeletal muscle injury samples across species into a machine learning model, cross-species predictions of the time points of human skeletal muscle injury can be obtained.

Images