SMR Analysis Integrating GWAS and eQTL Data Reveals UHRF1BP1 and SNRPC as Potential Drug Targets for Low Back Pain

Background

As a common clinical complaint, low back pain (LBP) can be clinically classified into three main types: overt injury pain (eg., myofascial LBP), neuropathic pain (eg., lumbar radiculopathy), and centrally sensitized pain (nonspecific LBP),^1,2 each potentially involving distinct molecular pathways and thus requiring different management strategies. The World Health Organization’s Global Burden of Disease Study shows that LBP has become the leading cause of disability in disability-adjusted life years (DALYs) in 126 countries,³ with a one-month prevalence rate of 23.3%,⁴ lifetime prevalence is as high as 47%,⁵ resulting in a loss of social productivity that exceeds that of most chronic diseases. Notably, the development of chronic LBP involves a multidimensional interaction of biomechanical abnormalities, altered neuroplasticity, and psychosocial factors,⁶ a complexity that often makes it difficult to achieve sustained relief with existing treatment options.

Current clinical guidelines recommend a stepwise therapeutic strategy: non-steroidal anti-inflammatory drugs (NSAIDs) are preferred for mild-to-moderate pain in the acute phase, and weak opioids can be used in combination with NSAIDs for moderate-to-severe pain; for chronic skeletal and musculoskeletal pain, central neuromodulators such as duloxetine can inhibit pain signaling by modulating 5-hydroxytryptamine/norepinephrine reuptake.^7–9 However, the limitations of insufficient response to conventional medications in partial patients, the potential for gastrointestinal damage¹⁰ and cardiovascular disease risk¹¹ with long-term use of NSAIDs, the risk of addiction with long-term opioid use highlight the urgent need for the development of novel targeted therapeutic regimens.

The traditional drug discovery model relies on phenotypic screening and randomized controlled trials, which is characterized by low target discovery efficiency and long development cycle. Genomics-based integration strategy provides a new direction to break through this dilemma: by integrating genome-wide association studies (GWAS) and expression quantitative trait loci (eQTL) data, the causal chain between genetic variation, gene expression and disease phenotype can be systematically analyzed.^12,13 Mendelian randomization (MR) uses genetic variants as instrumental variables to infer causality, reducing bias from confounding and reverse causation.^14,15 The crucial step in MR is selecting instrumental variables that satisfy three assumptions: (i) the instrumental variable should be closely related to the exposure of interest; (ii) the instrumental variable should be independent of the outcome of interest; (iii) the instrumental variable should be independent of all confounding factors related to the relationship between exposure and outcome.

Among MR approaches, summary-data-based Mendelian randomization (SMR) is particularly effective in detecting gene–trait associations, as it integrates GWAS and eQTL summary data and incorporates a heterogeneity in dependent instruments (HEIDI) test to distinguish pleiotropy from linkage.¹⁶

Therefore, inspired by the successful experience of SMR in Parkinson’s disease,^17–19 Alzheimer’s disease,^20–22 as well as in other pain conditions such as migraine,^23,24 we applied cross-tissue SMR (MR based on summarized data) integrating GWAS with blood and brain eQTL data, combined with pathway, PheWAS, and pharmacogenomic resources, to identify and prioritize druggable gene targets for LBP.

Methods

Study Design

Figure 1 illustrates the analytical framework of this study based on cross-omics Mendelian randomization, and the specific process includes the following key steps:

Figure 1 Overview of the study design. Abbreviations: eQTL, expression quantitative trait loci; SMR, Summary-data-based Mendelian randomization; HEIDI, heterogeneity in dependent instruments; FDR, False Discovery Rates.

(i) multi-tissue eQTL data integration: the peripheral blood eQTL dataset²⁵ constructed by Westra et al (sample size = 5,311) and the brain tissue eQTL²⁶ atlas of the United Kingdom Brain Expression Consortium (UKBEC) (covering 10 brain regions) were selected as the sources of instrumental variables, which systematically covered the peripheral and central gene regulatory networks related to pain signaling.

(ii) causal effect inference: based on the summarized GWAS data for LBP released by the FinnGen Consortium (version R11), gene-phenotype association analysis was performed using the Mendelian randomization method based on the summarized data, and multiple tests were rigorously corrected using the Bonferroni correction and FDR (False Discovery Rates) correction.

(iii) verification of genetic homogeneity: in order to exclude the interference of pleiotropy on the results, gene loci with a single causal pathway with the target phenotypes were screened by heterogeneity in dependent instruments (HEIDI) analysis (HEIDI P > 0.05) to ensure the genetic association specificity of the candidate target genes.

(iv) molecular pathway mining: based on the Reactome and GeneMania platforms, the pathways and interactions of the screened target genes will be analyzed.

(v) clinical translational assessment: Phenome-wide association scanning (PheWAS) was performed on validated targets to assess potential off-target effects, and pharmacological matching analysis was carried out based on the Drug Signatures Database (DsigDB) to predict repositionable drugs and their mechanisms of action, and finally, based on the Open targets and PhenomeXcan platforms, a further comprehensive assessment of the association of the target genes for LBP with other human traits or diseases.

Exposure Data

Table S1 details the multi-tissue eQTL data sources used in this study and their quality control processes:

Peripheral Blood eQTL Dataset

Based on the cross-cohort meta-analysis published by Westra’s team (5,311 whole blood samples from healthy individuals were included), a total of 19,725 protein-coding genes were examined for their expression profiles using a unified RNA sequencing process and a standardized genotyping platform. The screening criteria in this study included (i) cis-acting eQTL (cis-eQTL, defined within ±1 Mb from the gene transcription start site); (ii) genome-wide significant association signal (P < 5×10⁻⁸); (iii) independent locus screening (r² < 0.001 by PLINK clumping algorithm, window size 10 Mb). A final set of 15,636 gene probes meeting the above criteria was included. The raw data are publicly available through the European Bioinformatics Institute (EBI) database (accession number: EGAS00001000910).

Brain Tissue eQTL Dataset

Raw data derives from the UK Brain Expression Consortium (UKBEC) multi-brain region exome study, covering 134 individuals of European descent without a history of neurological disease (postmortem autopsy samples) and covering 10 pain-related brain regions such as prefrontal cortex and hippocampus. Gene expression quantification was performed using Illumina HumanHT-12v4 microarrays, and confounders such as age, gender and postmortem interval were corrected by mixed linear modeling. Strict screening conditions included (i) cis-acting eQTL (defined as above); (ii) genome-wide significance threshold (P < 5×10⁻⁸); (iii) probe-specific filtering (removal of cross-hybridization risk probes). 16,309 high-confidence probes were retained after QC, and data access is available on the consortium website (http://www.ukbec.org).

Outcome Data

Low back pain

The outcome dataset is from FinnGen version 11 and is available at https://r11.finngen.fi. LBP was defined using FinnGen endpoint codes M13_LOWBACKPAIN (inclusion of ICD-10 M54.4-M54.5 and corresponding ICD-8/9 codes). In this dataset, the total number of LBP was 13,178, while 164,682 cases were categorized as healthy.

Statistical Analysis

SMR Analysis and HEIDI Testing

In contrast to most other methods used for comprehensive analysis of GWAS and eQTL data, the SMR and HEIDI methods were able to distinguish between multiple and chained models.¹⁶ SMR analysis was performed using SMR software as a means of confirming causal relationships between LBP and gene expression.²⁷ Analyses were performed using SMR v1.3.1 with default parameters, using the European 1000 Genomes reference panel for LD clumping. R v4.2.0 was used for secondary analyses.

The HEIDI test was used to demonstrate that proteins associated with LBP were not caused by genetic linkage. A significant SMR association was defined as P < 0.05, whereas a HEIDI P > 0.05 indicated that the association was caused by shared genetic variation. Sensitivity analyses were then performed using both Bonferroni correction and FDR correction (Q-value) for significance thresholds. Peripheral blood P < 3.2×10⁻⁶ and brain P < 3.07×10⁻⁶ was considered significant in the Bonferroni correction. Q < 0.05 was considered significant in the FDR correction. The study was conducted according to current guidelines.

Tissue Expression Analysis

Human Protein Atlas (www.proteinatlas.org) was used to examine mRNA and protein expression across 44 normal human tissues cover 16,975 genes.²⁸ Exploring the expression of target genes in different tissues and different brain regions could suggest possible mechanisms for their use as therapeutic targets for LBP.

Pathway Analysis

Reactome (https://reactome.org) and GeneMANIA²⁹ (https://genemania.org) were used for functional enrichment and network analysis.

Drug Prediction

DSigDB³⁰(v1.0, http://dsigdb.tanlab.org) via Enrichr³¹ (https://maayanlab.cloud/modEnrichr/) was used to identify potential drug candidates.

PheWAS

AstraZeneca PheWAS Portal³²(https://azphewas.com) and PheWeb were used to assess pleiotropic effects and potential off-target risks.³³

Cross-Trait Analysis

Open Targets (https://www.targetvalidation.org) and PhenomeXcan³⁴ were used to evaluate associations with other human traits and diseases.

Results

SMR and HEIDI Test for the Association Between Peripheral Blood Gene Expression and LBP

Using FDR correction (Q < 0.05), SMR analysis of peripheral blood eQTL data identified numerous genes significantly associated with LBP risk (Figure 2A and Table S2). Under the more stringent Bonferroni correction (P_SMR < 3.2×10⁻⁶), only four genes remained significant: BTN2A3P (OR (95% CI) = 1.733 (1.425–2.109); P = 2.70×10⁻⁸), GFPT1 (OR (95% CI) = 1.094 (1.052–1.138); P = 1.34×10⁻⁷), UHRF1BP1 (OR (95% CI)= 1.051 (1.031–1.072); P = 6.56×10⁻⁷), SNRPC (OR (95% CI)= 0.869 (0.820–0.922); P = 8.01×10⁻⁷). All four passed HEIDI test (P > 0.05), supporting a causal role independent of linkage (Figure 2C and Table S2).

Figure 2 Manhattan plot for correlation of target genes with LBP in SMR and HEIDI test The y-axis shows –log10(P) or –log10(Q) values from SMR tests. The red horizontal line indicates the genome‑wide significance threshold for each panel. (A) Causal associations with LBP identified in peripheral blood eQTLs using FDR correction (FDR-P =0.05). (B) Causal associations with LBP identified in peripheral blood eQTLs using Bonferroni correction (PSMR = 3.2×10−6 = 0.05/15,636). (C) Causal associations with LBP identified in brain eQTLs using FDR correction (FDR-P = 0.05). (D) Causal associations with LBP identified in brain eQTLs using Bonferroni correction (PSMR =3.07×10−6 = 0.05/16,309). Genes passing the respective threshold are labeled.

SMR and HEIDI Test for the Association Between Brain Gene Expression and LBP

Similarly, in brain eQTL data, FDR correction identified a large set of genes (Figure 2B). After Bonferroni correction (P_SMR < 3.07×10⁻⁶), only four genes remained significant: CHST3 (OR (95% CI) = 0.923 (0.905–0.941); P = 6.87×10⁻⁹), DCC (OR (95% CI) = 1.174 (1.106–1.245); P = 3.63×10⁻⁸), UHRF1BP1(OR (95% CI) = 1.051 (1.031–1.072); P = 1.21×10⁻⁶), SNRPC (OR (95% CI) = 0.942 (0.923–0.960); P = 2. 13×10⁻⁶). All four passed HEIDI test (P > 0.05) (Figure 2D and Table S3).

UHRF1BP1 and SNRPC are the Concordant Genes Across Peripheral Blood and Brain eQTL Analyses

Notably, when comparing the Bonferroni-significant genes from the two tissues, UHRF1BP1 and SNRPC were the only genes that appeared in both lists (Figure 3C). Although other genes (eg., BTN2A3P, GFPT1 in blood; CHST3, DCC in brain) passed Bonferroni correction in a single tissue, their effects were tissue-specific. In contrast, UHRF1BP1 and SNRPC showed consistent causal associations with LBP in both peripheral blood and brain, and they also passed the stringent Bonferroni correction in each tissue. This cross-tissue consistency strongly suggests that these two genes are the most robustly prioritized candidate targets. The two genes that were highly concordant between the SMR results in peripheral blood eQTL data (Figure 3A) and brain eQTL data (Figure 3B) were UHRF1BP1 and SNRPC. This indicates the existence of a single potentially causal variant with pleiotropic effects on the three “phenotypes”: gene expression levels in blood, expression levels in brain, and LBP risk.

Figure 3 The causal relationship between peripheral blood genes and brain genes with the risk of LBP (A) The forest plot shows significant causal associations between four peripheral blood genes and the risk of LBP; (B) The forest plot shows significant causal associations between four brain genes and the risk of LBP; The effect estimates represent the odds ratio per SD increase of target genes on LBP, and the error bars represent 95% CIs. A P value of <0.05 was considered significant; (C)Venn diagram of target genes from two datasets.

Taken together, these findings highlight the clinical relevance of UHRF1BP1 and SNRPC as genetically supported candidate targets for LBP, warranting further investigation in therapeutic development.

Expression of UHRF1BP1 and SNRPC in Different Tissues

The differences in the expression of UHRF1BP1 and SNRPC in human tissues and brain regions by Human Protein Atlas were determined. The results showed that UHRF1BP1 was mainly expressed in skeletal muscle, bone marrow and cerebellum, cerebral cortex, and choroid plexus (Figure 4A and B). SNRPC was mainly expressed in cardiac muscle, bone marrow, skeletal muscle and spinal cord, white matter, and choroid plexus (Figure 4C and D). This may provide an explanation for the target tissues and pathways of UHRF1BP1 and SNRPC.

Figure 4 The expression of UHRF1BP1 and SNRPC in human tissues and brain regions (A) Expression of UHRF1BP1 in human tissues. (B) Expression of UHRF1BP1 in brain regions. (C) Expression of SNRPC in human tissues. (D) Expression of SNRPC in brain regions.

Reactome

Figure 5 shows the results of Reactome pathway analysis, indicating that the up-regulated gene UHRF1BP1 is involved in the CREB3 signaling pathway and pathways such as stress response and stimulus response. The expression of the down-regulated gene SNRPC was mainly enriched through the apoptotic signaling pathway by activation of cysteinyl asparaginase, and the detailed data is shown in Table S4.

Figure 5 Pathways involved in the alteration of UHRF1BP1 and SNRPC for LBP based on Reactome database (A) Pathways associated with UHRF1BP1 (upregulated in relation to LBP risk). (B) Pathways associated with SNRPC (downregulated in relation to LBP risk). The x‑axis represents the gene ratio (or enrichment fraction), and the y‑axis lists the pathways. Color intensity indicates –log10(FDR), and dot size represents the number of genes enriched in each pathway.

GeneMANIA

The network processing data tool GeneMANIA was used to identify new genes that may co-express or physically interact with our targets. The inner circle in Figure 6 includes target genes that we have entered into the engine, while the outer circle indicates new targets that show co-expression or physical interaction with our targets. This analysis was able to identify the most important biological pathways (all FDR P values < 0.05), with the top-ranked pathway being involvement in the small nuclear ribonucleoprotein complex. SNRPA, SNRNP70 were identified in the biological pathway analysis as target genes interacting with target genes related to LBP (eg. SNRPC). With this strategy, SNRPA, SNRNP70 might also be hypothesized to serve as a candidate target for LBP treatment.

Figure 6 Pathways and genetic interactions mainly related to the UHRF1BP1 and SNRPC based on GeneMANIA database.

DsigDB

Evaluation of UHRF1BP1 and SNRPC as possible drug targets required assessment of protein-drug interactions. To find possible drugs, UHRF1BP1 and SNRPC were examined using the DSigDB drug database on Enrichr. Based on the results, the top four drugs associated with UHRF1BP1 were levamisole (CTD00006205), dihydroquercetin (CTD00000115), nickel (CTD00006389), and hexestrol (CTD 00005818) (Table 1). The first four drugs associated with SNRPC queried in the DsigDB database all downregulate SNRPC expression, which increases the risk of developing LBP, and therefore these four drugs were not considered.

Table 1 Candidate Drug Predicted Using DSigDB

PheWAS

By Using the PheWAS Portal and PheWeb databases, phenotypic group-wide MR was performed to identify possible side effects of targeting UHRF1BP1 and SNRPC. The results revealed evidence of significant associations between UHRF1BP1, SNRPC and other phenotypes, ie., a strong association with hypertension, celiac disease, in either the PheWeb database or the PheWAS portal at the genome-wide significance level (P < 5×10⁻⁸) (Figure 7 and Table S5). These results reinforce the comprehensiveness of our findings, suggesting that if targeting UHRF1BP1, SNRPC is therapeutic, there is a low risk of adverse drug reactions or unexpected horizontal pleiotropic effects, but with attention to hypertension and celiac disease.

Figure 7 PheWAS analysis of off‑target risk for UHRF1BP1 and SNRPC based on the PheWeb database (A) Strong association between UHRF1BP1 and hypertension. (B) Strong association between UHRF1BP1 and celiac disease. (C) Strong association between SNRPC and hypertension. (D) Strong association between SNRPC and celiac disease. The dashed horizontal line indicates the genome‑wide significance threshold (P = 5×10−8).

Open Targets

Because many target genes are associated with other phenotypes, we aimed to determine in a more comprehensive manner which disease characteristics are affected by these exposures. A systematic analysis of potential diseases or phenotypes associated with target genes in the Open Target Platform was performed. Table 2 shows the phenotypes closely associated with UHRF1BP1 (also known as BLTP3A) and SNRPC, in which globalScore is the total association score and otGeneticsPortal (Open Target Genetics Portal) is the genetic association score. Phenotypes strongly associated with UHRF1BP1 were height, HDL cholesterol measurements, and body mass index. Phenotypes closely associated with SNRPC were LDL cholesterol assay, body mass index and platelet index. More phenotypes associated with UHRF1BP1 and SNRPC is detailed in Table S6.

Table 2 Phenotypes Closely Related to UHRF1BP1 and SNRPC Based on Open Targets

PhenomeXcan

In order to more comprehensively assess the association of LBP target genes with other phenotypes, PhenomeXcan was also applied to further analyze the unearthed target genes. Among them, RCP > 0.1 was defined as the causative gene. Table 3 shows that UHRF1BP1 is a risk gene for 18 phenotypes, of which the most closely related phenotypes are trunk fat mass, body fat percentage and waist circumference, and the detailed associations are shown in Table S7.

Table 3 Phenotypes Closely Related to UHRF1BP1 Based on PhenomeXcan

Discussion

As a global public health challenge, LBP is causing increasing physical and mental health damage and socioeconomic burden.³⁵ We identified UHRF1BP1 and SNRPC as genetically supported, mechanistically plausible candidate targets for LBP, through SMR-HEIDI integration of GWAS and multi-tissue eQTL data. Notably, the cross-tissue convergence of these two genes represents a more robust prioritization strategy than single-tissue or locus-based approaches.

The pathogenesis of LBP involves multi-system interactions, including local inflammatory cascades, dorsal root ganglion sensitization, and central nervous system remodeling. Among them, intervertebral disc degeneration (IVDD), as a key pathological basis,³⁶ has a central mechanism focusing on the imbalance of cellular homeostasis in the nucleus pulposus.³⁷ As the biomechanical core of the intervertebral disc, the nucleus pulposus (NP) has an extracellular matrix (ECM) composed mainly of type II collagen and proteoglycan. Human studies have shown that apoptosis in the nucleus pulposus not only leads to a decrease in ECM synthesis, but also activates neuroimmune interactions through the release of damage-associated molecular patterns (DAMPs),³⁸ a finding that echoes Yang et al’s observation that apoptosis inhibition ameliorates intervertebral disc degeneration in a rat model.³⁹ Therefore, targeted modulation of nucleus pulposus cell survival and inflammatory microenvironment may become an important strategy for IVDD intervention.

SNRPC encodes a core component of the U1 small nuclear ribonucleoprotein (snRNP) complex, which plays a fundamental role in pre-mRNA splicing and, more recently, has been implicated in neuronal gene regulation. Although direct evidence linking U1 snRNP or its core components to neuropathic pain or central sensitization is currently lacking, converging lines of evidence support the plausibility of such a connection. First, U1 snRNP dysfunction has been shown to cause RNA splicing defects that lead to neuronal hyperexcitability in neurodegenerative models,⁴⁰ establishing a precedent that splicing dysregulation can directly affect neuronal function. Second, the U1 snRNP core protein SNRNP70 (U1-70K) has been detected in axonal RNA-associated granules, suggesting that snRNP components can function locally in neuronal processes to regulate the axonal transcriptome and influence neural connectivity.⁴¹ Third, alternative splicing regulation has been directly implicated in neuropathic pain pathogenesis through other splicing factors,⁴² highlighting that the splicing machinery as a whole is relevant to pain processing. Given the high expression of SNRPC in the spinal cord and white matter observed in this study, we hypothesize that SNRPC may influence pain signaling in LBP through alternative splicing of genes involved in neuronal excitability, neuroinflammation, or apoptotic cascades—a hypothesis that requires experimental validation in future studies.

UHRF1BP1 (also known as BLTP3A) encodes a bridge-like lipid transfer protein involved in lipid transport and metabolism regulation.⁴³ Emerging evidence has established a strong link between lipid metabolism dysregulation and intervertebral disc degeneration (IDD), the primary pathological basis of LBP. Case-control and Mendelian randomization studies have demonstrated that dyslipidemia—particularly elevated triglycerides and apolipoprotein ratios—is significantly associated with increased risk of IDD and LBP.^44,45 Mechanistically, oxidized low-density lipoprotein (ox-LDL) drives endplate chondrocyte senescence and calcification via the ROS/P38-MAPK/NF-κB axis,⁴⁶ while excessive palmitic acid accumulation in nucleus pulposus cells induces endoplasmic reticulum stress-dependent senescence.⁴⁷ These findings directly link lipid dysregulation to key pathological processes in the degenerating disc, supporting the hypothesis that UHRF1BP1 may contribute to LBP through its role in lipid metabolism. Notably, this study found that UHRF1BP1 is highly expressed in the choroid plexus and cerebral cortex, suggesting potential involvement in central pain processing. Mechanistically, UHRF1BP1 binds to UHRF1, which has been shown to recruit histone deacetylase 1 (HDAC1) via its SRA domain to mediate histone deacetylation,⁴⁸ providing a molecular basis for UHRF1BP1 to potentially influence epigenetic regulation through the UHRF1-HDAC axis. HDACs have been well-established as key regulators of central sensitization and neuropathic pain: spinal HDAC upregulation induces nociceptive sensitization by inhibiting the GABAergic system;⁴⁹ HDAC6 inhibition alleviates neuropathic pain by modulating spinal STAT3 acetylation and downstream chemokine release;⁵⁰ and HDACs in the spinal cord and dorsal root ganglia regulate neuronal excitability and neuroinflammation through histone and non-histone deacetylation.⁵¹ Given the abundant HDAC evidence in CNS pain modulation, together with the cerebral expression of UHRF1BP1 and its indirect link to HDACs, we hypothesize that UHRF1BP1 may also participate in LBP through HDAC-mediated epigenetic regulation of central sensitization pathways. Together, we propose a dual-mechanism hypothesis: UHRF1BP1 may influence LBP pathogenesis through peripheral lipid-mediated disc degeneration and, simultaneously, through central HDAC-related epigenetic modulation of pain sensitization—both of which require direct functional validation in future studies.

Notably, PheWAS analysis showed that UHRF1BP1/SNRPC had a low risk of pleiotropy, and the DsigDB database suggested that natural compounds such as TAXIFOLIN might exert neuroprotective effects by modulating UHRF1BP1, which provided a basis for lead compound selection for translational medicine research. Multi-tissue eQTL integration analysis based on the SMR framework further strengthened the biological rationality of target selection: the strong expression signal of SNRPC in the spinal cord was highly consistent with its potential function of regulating peripheral sensitization, while the specific expression of UHRF1BP1 in the choroid plexus might be involved in the process of central sensitization by influencing the cerebrospinal fluid metabolism.

Several limitations should be acknowledged. First, the GWAS data are from European populations only, limiting generalizability. Second, the brain eQTL sample size (n=134) is modest, which may affect statistical power. Third, our analyses focus on transcript levels; proteomic and single-cell data are needed to resolve cell-type specific mechanisms. Fourth, the mechanistic interpretations for SNRPC and UHRF1BP1 are hypotheses generated from expression patterns and literature, not direct experimental evidence. Finally, LBP is clinically heterogeneous (radicular vs. mechanical), and stratified analyses were not possible.

Several translational challenges should be noted. First, the modest effect size (OR ≈ 1.05) implies that targeting a single gene may have limited impact, supporting the need for combination approaches—including psychological,⁶ physical, and muscle-strengthening interventions⁵²—as important adjuncts to pharmacotherapy. Second, the safety associations with hypertension and celiac disease, although not strong, warrant monitoring in future trials. Third, genetic findings reflect lifelong effects, whereas drugs act over shorter periods; dose-effect and duration studies are needed.

In conclusion, this integrative genomics study prioritizes UHRF1BP1 and SNRPC as genetically supported candidate therapeutic targets for LBP. These findings are hypothesis-generating and require experimental validation in cellular and animal models, as well as dose-effect studies to bridge the gap between genetic association and clinical translation.