Welcome to ACPred-LAF server

The discrimination of anti-cancer peptides (ACPs) is an important approach to benefit the development of ACPs through increasing productivity, shortening the development cycle, and reducing costs. Recently, many low-cost and effective predictors have been developed to identify ACPs. However, studies in this field have been suffered from feature engineering until now, which restricts the representation ability of the model to a certain extent.

In this study, we propose a novel deep-learning-based predictor named ACPred-LAF (Anti-Cancer peptide Predictor with Learnable and Adaptive Features based on sequence information and self-attention mechanism). In this predictor, we propose novel multi-sense and multi-scaled embedding methods to automatically learn and extract context sequential characteristics of ACPs. Benchmarking comparison results demonstrate that our ACPred-LAF performs better than the state-of-the-art methods both on existing benchmark datasets and our newly constructed dataset. Through the feature comparative analysis, we demonstrate that our learnable and self-adaptive embedding features are better than hand-crafted statistics features to capture the discriminative information, showing the great potential of widespread applications in related fields. The data interference experiment also proves the robustness of the model. Moreover, data annotation conflict appears in some benchmark datasets that were used in previous research, which might result in the potential evaluation bias. Here, we construct a new ACP benchmark dataset named ACP-Mixed to help the further research, which we expect to be a golden standard benchmark dataset in this field.

ACPred-LAF: Learning adaptive features based on a novel deep learning architecture to improve the predictive performance and robustness of anticancer peptides

March, 31st, 2021: the first version of ACPred-LAF server was established.