Instituto de Farmacología y Morfofisiología, Facultad de Ciencias Veterinarias, Universidad Austral de Chile, Valdivia 5110566, Chile.
Departamento de Ciencias Clínicas, Facultad de Ciencias Veterinarias y Pecuarias, Universidad de Chile, Santiago 8820808, Chile.
Int J Mol Sci. 2023 Jun 3;24(11):9705. doi: 10.3390/ijms24119705.
Drug combinations are increasingly studied in the field of anticancer agents. Mathematical models, such as Loewe, Bliss, and HSA, are used to interpret drug combinations, while informatics tools help cancer researchers identify the most effective combinations. However, the different algorithms each software uses lead to results that do not always correlate. This study compared the performance of Combenefit (Ver. 2.021) and SynergyFinder (Ver. 3.6) in analyzing drug synergy by studying combinations involving non-steroidal analgesics (celecoxib and indomethacin) and antitumor drugs (carboplatin, gemcitabine, and vinorelbine) on two canine mammary tumor cell lines. The drugs were characterized, their optimal concentration-response ranges were determined, and nine concentrations of each drug were used to make combination matrices. Viability data were analyzed under the HSA, Loewe, and Bliss models. Celecoxib-based combinations showed the most consistent synergistic effect among software and reference models. Combination heatmaps revealed that Combenefit gave stronger synergy signals, while SynergyFinder produced better concentration-response fitting. When the average values of the combination matrices were compared, some combinations shifted from synergistic to antagonistic due to differences in the curve fitting. We also used a simulated dataset to normalize each software's synergy scores, finding that Combenefit tends to increase the distance between synergistic and antagonistic combinations. We conclude that concentration-response data fitting biases the direction of the combination (synergistic or antagonistic). In contrast, the scoring from each software increases the differences among synergistic or antagonistic combinations in Combenefit when compared to SynergyFinder. We strongly recommend using multiple reference models and reporting complete data analysis for synergy claiming in combination studies.
药物组合在抗肿瘤药物领域的研究日益增多。数学模型，如 Loewe、Bliss 和 HSA，用于解释药物组合，而信息学工具帮助癌症研究人员确定最有效的组合。然而，每种软件使用的不同算法导致结果并不总是相关。本研究通过研究非甾体类镇痛药（塞来昔布和吲哚美辛）和抗肿瘤药物（卡铂、吉西他滨和长春瑞滨）组合在两种犬乳腺肿瘤细胞系中的协同作用，比较了 Combenefit（Ver. 2.021）和 SynergyFinder（Ver. 3.6）在分析药物协同作用方面的性能。对药物进行了特征描述，确定了最佳浓度-反应范围，并使用每种药物的九个浓度来制作组合矩阵。在 HSA、Loewe 和 Bliss 模型下分析了活力数据。基于塞来昔布的组合在软件和参考模型之间表现出最一致的协同作用。组合热图显示，Combenefit 给出了更强的协同信号，而 SynergyFinder 产生了更好的浓度-反应拟合。当比较组合矩阵的平均值时，由于曲线拟合的差异，一些组合从协同作用转变为拮抗作用。我们还使用模拟数据集对每个软件的协同评分进行了归一化，发现 Combenefit 倾向于增加协同和拮抗组合之间的距离。我们得出结论，浓度-反应数据拟合会影响组合的方向（协同或拮抗）。相比之下，在 Combenefit 中，与 SynergyFinder 相比，每个软件的评分增加了协同或拮抗组合之间的差异。我们强烈建议在组合研究中使用多个参考模型并报告完整的协同作用数据分析。
