Predictive Analyses of Tumor Relapse-free Survival Prognosis Related to Consistency between Different Cancer Tissues...

Traditional drug discovery faces challenges, including high costs, time-intensive processes, and inherent risks. The disease-specific signature-based Connectivity Map (CMap) approach is widely utilized. However, the commonly employed method for constructing disease-specific signatures, known as Differentially Expressed Genes (DEGs), suffers from inconsistencies between dysregulated genes and their prognostic implications in tumor tissue, as well as discrepancies in prognosis genes between tumor and normal tissues. In this study, we present predictive analyses of CMap-based drug repurposing for solid tumors that account for discrepancies in gene activity between tumor and adjacent normal tissues.

We propose a predictive approach, Prognosis Consistency Scoring (PCS), designed to address these inconsistencies by analyzing six types of solid tumors: BRCA, HNSC, LIHC, LUAD, LUSC, and THCA. PCS measures the consistency of gene prognosis between tumor and normal tissues by integrating the Recurrence- Free Survival (RFS) prognostic power of genes in both contexts. Disease-specific signatures are then constructed based on PCS, and drug repurposing is performed using the CMap and Lincs Unified Environment (CLUE). The predicted drugs were validated using data from DrugBank, ClinicalTrials, CancerDrugs_DB, Re- DO_Trials_DB, GDSC2, CTRPv2, and PRISM. Biological enrichment analysis was conducted via the Metascape platform.

Our findings reveal that these inconsistencies are pervasive. Compared to signatures based on DEGs, PCS-based signatures exhibit superior performance, identifying more drugs with higher prediction accuracy, as confirmed by DrugBank annotations. Notably, a significant proportion of predicted drugs without corresponding indications were subsequently validated in the ClinicalTrials database. For instance, three trials support the efficacy of gemcitabine in LIHC, and two trials support the use of dasatinib for the treatment of LUAD and LUSC. The prediction accuracy of PCS is higher than that of DEGs (20/31 vs. 1/31, Fisher's exact test, p-value < 0.001). In the CancerDrugs_DB and ReDO_Trials_DB databases, the performance of PCS was also influenced by DEGs. In the three in vitro tumor cell line drug response databases, including GDSC2, CTRPv2, and PRISM, PCS outperformed DEGs. Comparative analysis between PCS and Tumor Prognosis Score (TPS) indicates that, although TPS is a component of PCS, the performance of PCS is still influenced by TPS. Additionally, PCS-based signatures demonstrated elevated disease specificity and association with Drug-Related Genes (DRGs).

New drug discovery is time-consuming and risky. To address this, we proposed PCS, a method that improves drug repurposing accuracy by considering consistency between gene dysregulation and prognosis in tumor and normal tissues. PCS outperformed DEG-based approaches in predicting drugs and showed higher validation in DrugBank and ClinicalTrials. It also demonstrated disease specificity and supported rational combination therapies. PCS offers a more precise alternative and holds promise for further refinement.

PCS outperformed DEGs in drug repurposing predictions and exhibited higher disease specificity. The availability of tumor-normal paired samples with RFS was limited. Our method lacked comparisons with alternative approaches. PCS exhibited limited predictive utility in certain tumor types, suggesting its potential unsuitability for all solid tumors. PCS had a notable flaw, as positive scores also enriched some treatment drugs.