标题： 基于子序列重排序预训练的零样本学习用于化合物-蛋白质相互作用 显示英文标题

标题： Zero-Shot Learning with Subsequence Reordering Pretraining for Compound-Protein Interaction

摘要： 鉴于化学空间的广阔性和不断出现的未表征蛋白质，零样本化合物-蛋白质相互作用（CPI）预测更能反映现实世界药物开发的实际挑战和需求。 尽管现有方法在某些CPI任务中表现良好，但仍面临以下挑战：（1） 从局部或完整蛋白质序列中进行表示学习常常忽视子序列之间的复杂依赖关系，而这些关系对于预测空间结构和结合特性至关重要。 （2） 依赖于大规模或稀缺的多模态蛋白质数据集需要大量的训练数据和计算资源，限制了可扩展性和效率。 为了解决这些挑战，我们提出了一种新方法，通过子序列重新排序对蛋白质表示进行预训练，以显式捕捉蛋白质子序列之间的依赖关系。 此外，我们应用长度可变的蛋白质增强方法，以确保在小规模训练数据集上具有出色的预训练性能。 为了评估模型的有效性和零样本学习能力，我们将它与各种基线方法结合。 结果表明，我们的方法可以提高基线模型在CPI任务上的性能，尤其是在具有挑战性的零样本场景中。 与现有的预训练模型相比，我们的模型表现出更优越的性能，特别是在训练样本有限的数据稀缺场景中。 我们的实现可在 https://github.com/Hoch-Zhang/PSRP-CPI 获取。 显示英文摘要

摘要： Given the vastness of chemical space and the ongoing emergence of previously uncharacterized proteins, zero-shot compound-protein interaction (CPI) prediction better reflects the practical challenges and requirements of real-world drug development. Although existing methods perform adequately during certain CPI tasks, they still face the following challenges: (1) Representation learning from local or complete protein sequences often overlooks the complex interdependencies between subsequences, which are essential for predicting spatial structures and binding properties. (2) Dependence on large-scale or scarce multimodal protein datasets demands significant training data and computational resources, limiting scalability and efficiency. To address these challenges, we propose a novel approach that pretrains protein representations for CPI prediction tasks using subsequence reordering, explicitly capturing the dependencies between protein subsequences. Furthermore, we apply length-variable protein augmentation to ensure excellent pretraining performance on small training datasets. To evaluate the model's effectiveness and zero-shot learning ability, we combine it with various baseline methods. The results demonstrate that our approach can improve the baseline model's performance on the CPI task, especially in the challenging zero-shot scenario. Compared to existing pre-training models, our model demonstrates superior performance, particularly in data-scarce scenarios where training samples are limited. Our implementation is available at https://github.com/Hoch-Zhang/PSRP-CPI.