2.1 Patients

A total of 25 patients diagnosed with DDH (hereafter referred to as the DDH group) and 25 healthy controls (HC group) were enlisted from the People's Hospital of Tibet Autonomous Region between January 2021 and March 2022. The diagnosis was established by a seasoned clinician based on comprehensive examination findings. To be eligible for the study, participants had to have a verified diagnosis of one of these conditions: (1) a total hip joint dislocation, where there was no contact between the original joint surfaces; (2) a partial hip joint dislocation, with some contact remaining between the joint surfaces; or (3) acetabular dysplasia, marked by underdevelopment of the acetabulum. Experienced pediatric orthopedic experts made the diagnoses after conducting thorough medical assessments. Other clinical signs taken into account included the noticeable upward and outward movement of the greater trochanters, limited hip joint abduction, and positive Ortolani, Barlow, Allis, or Galeazzi signs. X-ray imaging was used to confirm the diagnoses. The control group was made up of healthy children in Tibet who had regular health check-ups at the People's Hospital of Tibet and showed no abnormalities.

Fasting blood samples were drawn using venipuncture into Vacuette tubes prefilled with a procoagulant. Within a frame of 15–30 min post-collection, the samples underwent centrifugation at 1200×g for a duration of 10 min. The clinical data and hospital records pertaining to all participants are delineated in Table 1. Biospecimens from patients were collected following approval from the People's Hospital of Tibet Autonomous Region institutional committee tasked with protecting human subjects (ME-TBHP-21-023). Informed consent was procured from each participant subsequent to thorough clarification regarding the procedures and intent of the study.

2.2 Materials and Equipment

Reagents such as mass spectrometry grade ammonium formate and dichloromethane were procured from Fisher Chemical, while mass spectrometry grade methanol and acetonitrile were sourced from CNW Technologies. For lipidomics profiling, we employed a liquid chromatography system (1290, Agilent Technologies). interfaced with the Q Exactive Focus (Thermo Fisher Scientific). Separation was achieved using the ACQUITY UPLC BEH Amide column (1.7 μm, 2.1 × 100 mm, Phenomen).

2.3 Sample Preparation

For the lipidomics assessment, 100 μL of each specimen was amalgamated with 480 μL of a methanol/MTBE solution (mixed in a 1:5 volume ratio). This amalgamation took place in a sterile EP tube. Subsequent to vigorous vortexing, the samples were sonicated for 10 min in an ice-water bath. Post incubation of 1 h at −40°C, the samples were subjected to centrifugation at 3000 rpm for 15 min at 4°C. The supernatant was then earmarked for mass spectrometry analysis. It should be noted that the quality control (QC) sample was constituted by pooling an equal volume 20 μL from all the individual samples.

2.4 Lipidomics Analysis

MS analyses were carried out using an UPLC system (1290, Agilent Technologies) coupled to an Q Exactive Focus (Thermo Fisher Scientific) operating in data-dependent acquisition (DDA) mode. The separation was achieved using the ACQUITY UPLC BEH Amide column (1.7 μm, 2.1 × 100 mm, Phenomen). The mobile phase comprised 10 mmol/L ammonium formate and 60% acetonitrile in water (pH = 9.75) (A), and 10% acetonitrile in isopropanol (B). The autosampler was maintained at 4°C with an injection volume set at 2 μL. ESI source conditions were as follows: capillary temperature at 320°C, spray voltage at 5 kV (positive) or −4.5 kV (negative), sheath gas flow rate at 30 Arb, auxiliary gas flow rate at 10 Arb, full MS resolution at 70,000, MS/MS resolution at 17,500, collision energy at 15/30/45 in NCE mode.

2.5 Data Processing

For lipidomics analysis, raw MS data were transformed with ProteoWizard software to mzXML. Initially, we conducted a range of data management tasks, which comprised the following steps: (1) filtering out off-values, (2) eliminating missing values, (3) imputing missing values, and (4) normalizing the data. In this process, the internal standard (IS) was employed for the normalization step. Selecting the ion mode is an important step in metabolomics, mainly based on the properties of compounds and the mobile phase environment. In our project, we used both modes. Tasks such as peak picking, peak extraction, alignment, and integration were accomplished using the CentWave algorithm in XCMS software, identification was accomplished using LipidBlast library.

2.6 Statistical Analysis

Data are presented as mean ± SD. Statistical evaluations were conducted using the R package (ropls). Within the orthogonal partial least-squares discriminant analysis (OPLS-DA) model, the variable importance in the projection (VIP) value for each parameter was computed. Metabolites with a VIP value exceeding 1 were further assessed for significance using the Student's t-test at the univariate level. A p-value of less than 0.05 was deemed statistically significant. The heatmap and receiver operating characteristic (ROC) analyses were generated utilizing MetaboAnalyst 5.0 (Xia Lab @ McGill, Sweden).

3.1 Clinical Data of DDH Patients

All participants in the study were of Tibetan ethnicity (100%). The average age of the children enrolled was 19.9 ± 9.1 months. The residence altitude was recorded as 4101.9 ± 344.7 meters. None of the participants had a family history of developmental dysplasia of the hip (DDH). The gender ratio was 18 girls to 7 boys. During the fetal period, there were two cases (8%) of oligohydramnios, one case (4%) of post-term birth, and all deliveries (92%) were vaginal with cephalic presentations. Four cases (16%) represented primiparous mothers. The average birth weight was 3.5 ± 0.5 kg, with nine infants (36%) weighing more than 4 kg at birth. Swaddling with tight wrappings, such as candle wraps, was practiced in eleven cases during infancy (44%). Among the total number of children evaluated, seven had left hip involvement, eight had right hip involvement, and eleven had bilateral hip involvement respectively. Open surgical treatment was performed in twenty-two cases (88%), following diagnosis (Table 1).

3.2 Raw Data Preprocessing

To mitigate the influence of detection system errors and enhance the biological relevance of our findings, we performed rigorous data management on the primary data set. This process encompassed several steps:

3.3 Cluster Analysis (OPLS-DA)

Given the high-dimensional nature of the metabolomic data (with numerous detected metabolites) and the limited sample size, we utilized OPLS-DA. This analytical method allowed us to exclude orthogonal variables in metabolites, which are not related to categorical distinctions, providing a more accurate depiction of variations among metabolite groups and the correlation magnitude of the test group.

In the provided figure, the horizontal axis t[1]P signifies the predicted primary component score of the first principal component, highlighting intergroup sample differences. The vertical axis t[1]O denotes the orthogonal primary component score, underscoring intra-group sample variations. The results from the OPLS-DA score plot indicate a significant distinction between the two sample groups, with most samples falling within a 95% confidence interval (Figure 1).

3.4 Volcano Map

The volcano map provides a visual representation of the overall distribution of differential metabolite levels between the groups. The two primary indicators: a p-value from the student's t-test less than 0.05 and a Variable Importance in the OPLS-DA Model greater than 1. In this map, each point symbolizes a peak. Metabolites that are significantly upregulated are depicted in red, downregulated metabolites are in blue, and those without significant differentiation are in gray (as shown in Figure 2).

3.5 Heatmap

The heatmap was constructed through the quantitative values of differential metabolites by calculating the Euclidean distance matrix. A complete linkage clustering method was employed to cluster these metabolites. In the heatmap, the x-axis represents the different experimental groups while the y-axis denotes the varied metabolites being compared within the groups. The color coding within the heatmap (as shown in Figure 3) corresponds to the relative expression levels of the metabolites, with red indicating higher expression and blue suggesting lower expression.

3.6 Matchstick Diagram and Radar Chart Analysis

We determined the corresponding ratio for the differential metabolites’ quantitative values, selecting the top 15 for both upregulation and downregulation to present the results. In the diagram, changes in multiple are illustrated on the x-axis, while the intensity of the dot color indicates the magnitude of the VIP value. This analysis highlights metabolites exhibiting substantial alterations, suggesting that the expression level of their associated enzymes could be either significantly inhibited or activated (*0.01 < p < 0.05, **0.001 < p < 0.01, ***p < 0.001). Further validation can be conducted to examine the regulatory impact on enzyme gene expression by certain metabolites pertinent to the study (as depicted in Figure 4).

For the radar chart, we again computed the corresponding ratio for the differential metabolite's quantitative values and applied a logarithmic transformation with a base of 2. In the chart, purple shadows are formed by the lines indicating the difference multiples for each metabolite, showcasing the evolving trend of content in the radar map (refer to Figure 5).

3.7 Bubble Chart

Our metabolic pathway analysis results are visually represented in a bubble chart. The bubble's horizontal size and position demonstrate the pathway's influence factor based on topological analysis. A darker bubble shade indicates a smaller p-value, denoting a higher enrichment significance. The main lipid species that showed differences include primary triacylglycerols (TAGs) and phosphatidylethanolamine (PE). In the field of bioinformatics, the area under the curve (AUC) is frequently utilized to assess the effectiveness of classification models. Ranging from 0 to 1, this metric gauges the model's capacity to categorize samples. Typically, the closer the AUC is to 1, the greater the reliability of the detection method. Conversely, an AUC of 0.5 indicates the lowest reliability and lacks practical value. In this study, we examined the AUC of various metabolites (see Figure 6, Table 2). TAG (17:0/18:1/20:1) and TAG (17:0/17:0/17:0) have the potential to become valuable diagnostic tools in the future.