About this Research Topic
Peptides play essential roles in human physiology and can be used to mimic the function of proteins in the body, making them attractive candidates for drug development. Peptide-based drugs are being developed in many therapeutic areas including cancer, immune disorders, cardiovascular diseases, gastrointestinal dysfunction, haemostasis and microbial infections. However, many peptides do not translate into the clinic due to their lack of metabolic stability, lability during storage, limited oral bioavailability and unwanted toxicity. Moreover, experiments in peptide drug design must balance between improving the biological activities (e.g., target specificity, binding affinity) and reducing undesirable side effects (e.g., aggregation, immunogenicity) of a given peptide– a multi-objective optimization problem. Consequently, the process of discovering and designing peptide-based drugs is complex and time-consuming.
Artificial intelligence (AI) has the potential to revolutionize the discovery and design of peptide-based drugs. AI algorithms learn patterns from large amounts of data, such as the prediction of biological activities (e.g., target binding affinity, hemotoxicity, etc.) from sequential or structural information. They save time and resources in the drug development process. These algorithms can also generate novel sequences from existing peptide families or folds. The combined use of predictive and generative models helps design peptides with specific properties, such as optimal stability and solubility, making them more effective potential drugs. Finally, AI can be used to optimize peptide-protein interactions, which can lead to the development of more potent drugs.
There are also limitations to the use of AI in the discovery and design of peptide-based drugs. Limitations in peptide predictive and generative modelling lie in the diversity of peptide sequences and biological information. Biases and imbalances in harnessing chemical and biological information impact the performance of machine learning models. Understanding the boundaries and biases of our existing AI-powered applications to peptide-based drug discovery is necessary for their applicability to the real-world.
In this Research Topic, we welcome submissions of Original Research, Review, Mini Review, Opinion, and Perspective articles, on the following, but not limited to, topics:
- The effects of peptide embeddings on predictive and generative models
- Machine learning models for peptides with non-natural amino acids and chemical modifications.
- Machine learning models applied to multi-objectivity and promiscuous functionality
- Novel approaches to detect peptides “within-distribution” (inliers) and “out-of-distribution” (outliers)
- Novel methods to mitigate data imbalance and algorithmic bias
Artificial intelligence (AI) has the potential to revolutionize the discovery and design of peptide-based drugs. AI algorithms learn patterns from large amounts of data, such as the prediction of biological activities (e.g., target binding affinity, hemotoxicity, etc.) from sequential or structural information. They save time and resources in the drug development process. These algorithms can also generate novel sequences from existing peptide families or folds. The combined use of predictive and generative models helps design peptides with specific properties, such as optimal stability and solubility, making them more effective potential drugs. Finally, AI can be used to optimize peptide-protein interactions, which can lead to the development of more potent drugs.
There are also limitations to the use of AI in the discovery and design of peptide-based drugs. Limitations in peptide predictive and generative modelling lie in the diversity of peptide sequences and biological information. Biases and imbalances in harnessing chemical and biological information impact the performance of machine learning models. Understanding the boundaries and biases of our existing AI-powered applications to peptide-based drug discovery is necessary for their applicability to the real-world.
In this Research Topic, we welcome submissions of Original Research, Review, Mini Review, Opinion, and Perspective articles, on the following, but not limited to, topics:
- The effects of peptide embeddings on predictive and generative models
- Machine learning models for peptides with non-natural amino acids and chemical modifications.
- Machine learning models applied to multi-objectivity and promiscuous functionality
- Novel approaches to detect peptides “within-distribution” (inliers) and “out-of-distribution” (outliers)
- Novel methods to mitigate data imbalance and algorithmic bias
Keywords: Peptide-based drugs, Artificial intelligence (AI), Predictive and generative models, Biological activities, Machine learning models, Peptide embeddings, Outlier detection, Algorithmic bias
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
