Jinhan Liu, Yibo Yang, Ruiying Lu, Piotr Piękos, Yimeng Chen, Peng Wang, Dandan Guo. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Plasma protein binding (PPB) is an important pharmacokinetic parameter. It is important to measure the PPB properties of drug molecules during drug development. However, in vivo or in vitro measurements are time-consuming. Therefore, in silico prediction methods are promising time-saving alternatives. This study presents a new deep learning model called enhancing plasma protein binding prediction (ePPBP) that merges molecular descriptors, molecular fingerprints and graph features for predicting PPB. The ePPBP currently has state-of-the-art (SOTA) performance, with an Rp2of 0.8663, an R2of 0.8630, an MAE of 0.0613 and an RMSE of 0.1041 on the test set. In addition, an ablation experiment demonstrated that different molecular representations can improve ePPBP performance. Next, an uncertainty estimation experiment was used to estimate the confidence when ePPBP was used to predict unknown chemicals. The MHFP distance and RDKFP similarity were selected as confidence indicators to determine whether the predictions were credible from ePPBP.

Search and recommendation (S&R) are the two most important scenarios in e-commerce. The majority of users typically interact with products in S&R scenarios, indicating the need and potential for joint modeling. Traditional multi-scenario models use shared parameters to learn the similarity of multiple tasks, and task-specific parameters to learn the divergence of individual tasks. This coarse-grained modeling approach does not effectively capture the differences between S&R scenarios. Furthermore, this approach does not sufficiently exploit the information across the global label space. These issues can result in the suboptimal performance of multi-scenario models in handling both S&R scenarios. To address these issues, we propose an effective and universal framework for Unified Search and Recommendation (USR), designed with S&R Views User Interest Extractor Layer (IE) and S&R Views Feature Generator Layer (FG) to separately generate user interests and scenario-agnostic feature representations for S&R. Next, we introduce a Global Label Space Multi-Task Layer (GLMT) that uses global labels as supervised signals of auxiliary tasks and jointly models the main task and auxiliary tasks using conditional probability. Extensive experimental evaluations on real-world industrial datasets show that USR can be applied to various multi-scenario models and significantly improve their performance. Online A/B testing also indicates substantial performance gains across multiple metrics. Currently, USR has been successfully deployed in the 7Fresh App.