Integrated bulk and single-cell RNA sequencing to identify potential biomarkers in intervertebral disc degeneration.

Abstract

Background: Nucleus pulposus (NP) deterioration plays a significant role in the development of intervertebral disc degeneration (IVDD) and low back pain (LBP). This paper aims to identify potential genes within degenerated NP tissue and elucidate the pathogenesis of IVDD through bioinformatics analysis.

Methods: We conducted a transcriptomic analysis of patient's degenerative NP tissue employing advanced bioinformatics techniques and machine learning algorithms. Utilizing hdWGCNA, we successfully acquired WGCNA single-cell sequencing data and pinpointed crucial genes implicated in IVDD. Subsequently, we employed the Monocle3 package to perform pseudotime sequence analysis, enabling the identification of genes associated with the differentiation and developmental processes of NP tissue. Following this, normalized and logarithmically transformed the bulk sequencing data. Subsequently, we conducted preliminary screening using single-factor logistic regression on the genes derived from single-cell sequencing. Next, we applied two machine learning techniques, namely, SVM-RFE and random forest, to discern pivotal pathogenic genes. Finally, we used validation sets to verify trends and qualitativeness and performed in vitro and in vivo validation analyses of normal and degenerative NP tissues.

Results: 909 genes associated with IVDD were identified through hdWGCNA, while pseudotime sequence analysis uncovered 1964 genes related to differentiation and developmental processes. The two had 208 genes in common. Subsequently, we conducted an initial screening of single-cell genes by integrating the bulk database with single logistic regression. Next, we utilized machine learning techniques to identify the IVDD genes CDH, DPH5, and SELENOF. PCR analysis confirmed that the expression of CDH and DPH5 in degraded nucleus pulposus cells (NPCs) was decreased by 31% and 28% in vivo, and 36% and 29% in vitro, respectively, while SELENOF showed the opposite trend. Furthermore, IVDD was validated through imaging and histological staining.

Conclusion: As pathogenic genes in IVDD, our findings indicate that CTH, DPH5, and SELENOF are important players and might be promising therapeutic targets for IVDD treatment.