The investigators combined transcriptomic, proteomic, immunohistochemical, survival, functional-network and pharmacogenomic data from public resources, including The Cancer Genome Atlas, CPTAC/UALCAN, the Human Protein Atlas, GeneMANIA, STRING, GSCA and CPADS. After surveying SRPK1 expression across TCGA tumour types, they focused on breast invasive carcinoma, colon adenocarcinoma and prostate adenocarcinoma, abbreviated BRCA, COAD and PRAD. Bulk RNA-sequencing and proteomic measurements were combined with patient survival records, cancer-cell-line drug-response measurements and model-based drug-response estimates for TCGA tumour samples. All components were observational or computational; the investigators did not intervene on SRPK1 in patients or experimentally establish a mechanism.

The pan-cancer transcriptomic survey identified elevated SRPK1 messenger RNA in primary bladder, breast, cervical, cholangiocarcinoma, colon, esophageal, head-and-neck, liver, lung, stomach, read and uterine cancers, while expression was lower in kidney chromophobe and clear-cell renal carcinoma. No significant tumour-normal difference was detected in glioblastoma, kidney renal papillary carcinoma, pancreatic adenocarcinoma, prostate adenocarcinoma or thyroid carcinoma. Comparisons using combined TCGA and GTEx normal-tissue data confirmed higher SRPK1 expression in BRCA and COAD, but not PRAD, for which the difference was not significant. CPTAC proteomic data likewise showed higher SRPK1 protein in BRCA and COAD tumours than in adjacent normal tissue, both at p<0.001. In prostate cancer, the comparison was between age-stratified tumour groups rather than tumour and normal tissue; protein levels did not differ significantly between ages 41–60 and 61–80 years.

Human Protein Atlas immunohistochemistry offered a complementary view, showing predominantly cytoplasmic SRPK1 staining in tumour tissues from all three cancers and higher protein expression than in corresponding normal tissue. The prostate findings nevertheless remained less certain because the transcriptomic comparison was not significant, the age-stratified protein comparison was not significant and the available sample sizes were limited. The evidence therefore supports tumour-specific dysregulation, with the strength and direction of the signal varying by molecular layer and dataset, rather than universal overexpression across the three cancers.

Initial survival analyses, which divided patients in each cancer at median SRPK1 expression, produced weak or non-significant associations. Larger Kaplan-Meier Plotter cohorts produced more statistically significant results using an automatically selected cutoff. In BRCA, high SRPK1 expression was associated with worse overall survival in 1,104 high-expression patients versus 775 low-expression patients, with a hazard ratio of 1.57 and log-rank p=6.8×10^-6. The same direction appeared for disease-free survival in 2,473 high-expression and 2,456 low-expression cases, with a hazard ratio of 1.66 and p<1×10^-16. These were associations between tumour expression and survival; they do not show that increasing SRPK1 causes disease progression or that inhibiting it would improve outcomes.

The colon-cancer association ran in the opposite direction. Among 402 patients with high SRPK1 expression and 659 with low expression, high expression was associated with longer overall survival, giving a hazard ratio of 0.75 and p=0.0083. High expression was also associated with longer disease-free survival in 926 high-expression and 410 low-expression patients, with a hazard ratio of 0.52 and p=1.1×10^-9. Thus, the analysis associated lower SRPK1 expression with poorer outcomes in COAD. The divergence indicates that the prognostic meaning of SRPK1 depends on tumour type and should not be extrapolated from breast cancer to colorectal cancer or prostate cancer.

Interaction and co-expression analyses supplied a possible biological framework without establishing causality. GeneMANIA placed SRPK1 in a network of genes associated with messenger RNA splicing and processing, chromatin organization, transcriptional regulation, proteolysis, DNA-damage responses, the cell cycle and immune signalling. Using experimentally validated STRING interactions, the researchers identified serine/arginine-rich splicing factors SRSF1, SRSF2, SRSF3, SRSF4, SRSF5, SRSF6, SRSF7 and SRSF9, together with the spliceosomal helicase DDX23 and the U1 small nuclear ribonucleoprotein component SNRNP70. These databases indicate functional proximity but do not demonstrate that a particular interaction occurs in patient tumours or that SRPK1 phosphorylation changes a specific splice variant after chemotherapy.

SRPK1 and SRSF1 messenger RNA levels were positively correlated in all three focused cancers: r=0.45 in BRCA, r=0.63 in COAD and r=0.68 in PRAD, with p<0.001 for each correlation. The co-expression pattern was compatible with the known capacity of SRPK kinases to regulate serine/arginine-rich splicing factors, but its biological and therapeutic consequences remain unresolved. In Kaplan-Meier Plotter cohorts, high SRSF1 expression was associated with longer disease-free survival in both BRCA and COAD, with hazard ratios of 0.64 and 0.66 and p-values of 5.5×10^-9 and 0.00045, respectively. The breast-cancer relationship differed from that for SRPK1, demonstrating that correlated molecules need not have identical prognostic effects.

Pharmacogenomic results extended the analysis beyond survival. The GDSC dataset contained sensitivity measurements for 251 compounds in 1,001 cell lines, and SRPK1 expression had statistically significant but weak correlations with responses to the 30 highest-ranked agents, ranging from r=-0.25 to r=0.25. Nine associations were positive, interpreted as greater relative resistance at higher SRPK1 expression, and 21 were negative, interpreted as greater relative sensitivity. In the independent CTRP dataset, covering 481 compounds in 960 cell lines, all 30 leading associations were negative and ranged from r=-0.31 to r=-0.38. These modest pan-cancer coefficients were spread across diverse cellular lineages and should not be interpreted as strong or clinically meaningful drug biomarkers.

The more specific CPADS results were still model-based. TCGA COAD samples with high SRPK1 expression had lower predicted half-maximal inhibitory concentrations for cisplatin than low-expression samples, consistent with greater predicted sensitivity, with a two-sided Wilcoxon rank-sum p=0.0065. In PRAD, high SRPK1 expression was associated with higher predicted docetaxel IC50 values, consistent with predicted resistance, at p=0.018. The corresponding BRCA comparison favored higher predicted cisplatin IC50 values in the high-expression group, but the difference was not statistically significant at p=0.064. Because these values were predicted from gene-expression profiles and cell-line pharmacogenomic data rather than measured in treated patients, they generate hypotheses for experimental testing rather than evidence that an individual patient will respond to cisplatin or docetaxel.

Several limitations constrain generalization. Bulk RNA sequencing averages signals across cancer, stromal, immune and other microenvironmental cells, potentially obscuring cell-specific splicing circuits. Some survival cohorts were small, treatment information was incomplete and expression cutoffs were selected differently across platforms. The pharmacogenomic resources also contain relatively few cell lines for some cancers and cannot represent every influence on drug response, including protein activity, epigenetic state, tumour heterogeneity and clinical factors. Most importantly, the study did not manipulate SRPK1, measure SRPK1-dependent phosphorylation in resistant models, resolve downstream isoform switching or test whether changing SRPK1 alters chemotherapy sensitivity. Larger independent cohorts and controlled cancer-cell models that jointly measure SRPK1 activity, SRSF1 phosphorylation, specific splice variants and chemotherapy phenotypes are needed to determine whether the observed associations reflect causal biology.

Subject of Research: Tumour-specific associations of SRPK1 expression with survival, splicing networks and predicted chemotherapy response

Article Title: Pan-cancer multi-omics profiling of SRPK1 identifies tumour-specific associations with prognosis and predicted chemotherapy response in breast, colon and prostate cancer

Article References: Duzgun, D., & Oltean, S. (2026). Pan-cancer multi-omics profiling of SRPK1 identifies tumour-specific associations with prognosis and predicted chemotherapy response in breast, colon and prostate cancer. Frontiers in Bioinformatics, 6, Article 1925354. https://doi.org/10.3389/fbinf.2026.1925354

Image Credits: AI Generated

DOI: 10.3389/fbinf.2026.1925354

Keywords: SRPK1, multi-omics, pan-cancer analysis, alternative splicing, chemotherapy response, breast cancer, colon cancer, prostate cancer, prognosis, pharmacogenomics, Pan-cancer, profiling