Screening for Biomarkers for Progression from Oral Leukoplakia to Oral Squamous Cell Carcinoma and Evaluation of Diagnostic Efficacy by Multiple Machine Learning Algorithms

The aim of the study is to identify key genes during the progression from oral leukoplakia (OL) to oral squamous cell carcinoma (OSCC) and predict effective diagnoses. Weighted gene co-expression network analysis (WGCNA) and differential expression analysis were performed to identify seven genes associated with the progression from OL to OSCC. Twelve machine learning algorithms including k-nearest neighbor (KNN), neural network (NNet), and extreme gradient boosting (XGBoost) were used to construct multi-gene models, which revealed that each model had good diagnostic efficacy. The functional mechanism or the pathways associated with these genes were evaluated using enrichment analysis, subtype clustering, and immune infiltration analysis. The enrichment analysis revealed that the genes enriched were associated with the cell cycle, cell division, and intracellular energy metabolism. The immunoassay results revealed that the genes primarily affected the infiltration of proliferating T cells and macrophage polarization. Finally, a nomogram and Kaplan-Meier survival analysis were used to predict the prognostic efficacy of key genes in OSCC patients. The results showed that genes could predict the prognosis of the patients, and patients in the high-risk group had a poor prognosis. Our study identified that the seven key genes, including DHX9, BCL2L12, RAD51, MELK, CDC6, ANLN, and KIF4A, were associated with the progression from OL to OSCC. These genes had good diagnostic efficacy and could be used as potential biomarkers for the prognosis of OSCC patients.

Unraveling novel survivin mRNA transcripts in cancer cells using an in-house developed targeted high-throughput sequencing approach

The human baculoviral IAP repeat containing 5 (BIRC5), also known as survivin, is a conserved member of the inhibitor of apoptosis protein (IAPs) family, which is normally expressed during embryonic and fetal development. Although the expression levels of survivin are low in terminally differentiated cells and/or tissues, they can be found notably increased in certain pathological conditions including malignant tumors. Conventional cloning and sequencing techniques have already confirmed that alternative splicing events of the survivin pre-mRNA result in five distinct alternative transcript variants. In the present study, however, we implemented an innovative, in-house developed, targeted DNA-seq assay to identify novel survivin alternative transcript variants with increased depth and coverage that high-throughput sequencing approaches offer. Bioinformatics analysis of the derived NGS datasets unveiled several novel splice junctions between annotated exons of survivin gene as well as the existence of a novel exon of 117 nt, spanning between the annotated exons 3 and 3B. Validation of the NGS findings with PCR-based assays, using variant-specific primers, led to the identification of fourteen novel survivin alternative splice variants (BIRC5 v.4 - v.17), which demonstrate wide expression profiles in a broad established panel of human cell lines. Although the presented novel findings provide a crystal-clear overview of the survivin mRNAs that are actually generated from the pre-mRNA, future studies should focus on the impending necessity of characterizing the biological function of all novel alternative transcript variants as well as the putative protein isoforms. Such studies will further contribute to our understanding of how the balance between survivin isoforms regulate malignant cell proliferation and apoptosis, providing novel diagnostic, prognostic and predictive biomarkers as well as therapeutic targets.

Identification of Prognostic Alternative Splicing Signature in Breast Carcinoma

Background

Increasing evidence indicated a close relationship between aberrant splicing variants and carcinoma, whereas comprehensive analysis of prognostic alternative splicing (AS) profiling in breast cancer (BRCA) is lacking and largely unknown.

Methods

RNA-seq data and corresponding clinical information of BRCA patients were obtained and integrated from The Cancer Genome Atlas (TCGA). Then SpliceSeq software was used to assess seven AS types and calculate the Percent Spliced In (PSI) value. Univariate followed by stepwise multivariate Cox regression analyses identified survival associated AS events and constructed the AS signature, which were further sent for enrichment analysis, respectively. Besides, the splicing correlation network was constructed. Additionally, nomogram incorporating AS signature and clinicopathological characteristics was developed and its efficacy was evaluated with respect to discrimination, calibration and clinical utility.

Results

A total of 45,421 AS events were detected, among which 3071 events were found associated with overall survival (OS) after strict filtering. Parent genes of these prognostic events were involved in BRCA-related processes including NF-kappaB and HIF-1 signaling pathway. Besides, the final prognostic signature built with 20 AS events performed well with an area under the curve (AUC) of receiver operating characteristic (ROC) curve up to 0.957 for 5 years. And gene set enrichment analysis (GSEA) also confirmed the candidate 20 AS events contributed to progression of BRCA. Moreover, the nomogram that incorporated 20-AS-event-based classifier, age, pathological stage and Her-2 status showed good calibration and moderate discrimination, with C-index of 0.883 (95% CI, 0.844–0.921). Decision curve analysis (DCA) confirmed more benefit was added to survival prediction with our nomogram, especially in 5 or 8 years with threshold probability up to 80%. Finally, splicing correlation network revealed an obvious regulatory pattern of prognostic splicing factors (SF) in BRCA.

Conclusion

This study provided a systematic portrait of survival-associated AS events involved in BRCA and further presented a AS-clinicopathological nomogram, which could be conveniently used to assist the individualized prediction of long-term survival probability for BRCA patients. And a series of bioinformatic analysis provided a promising perspective for further uncovering the underlying mechanisms of AS events and validating therapeutic targets for BRCA.