Traditionally, drugs produced by pharmaceutical companies require substantial investment from initial discovery to clinical trials; the process can easily extend beyond fifteen years until a tested therapeutic is released as a medication. While some portions of the drug discovery and development cannot be expedited, namely the clinical trial phase for obvious safety purposes, researchers have for years been searching for ways to optimize other parts of the process. One such area that this can be achieved in is initial drug discovery, or the large-scale search for bioactive molecules, such as antibodies, proteins, or peptides, that could play a role in the disease of focus. Conventional early drug identification techniques, such as High-Throughput Screening (HTS), allow for researchers to test enormous amounts of compounds against specific biological targets. While these techniques have been a breakthrough in drug development, there are certain drawbacks with the methodology, especially in regards to its costs, resource demands, and data management.
As with many other fields, artificial intelligence (AI) has made sweeping advances in a short period of time. Trained on various protein databases, these models represent a fresh effort to revolutionize computational capabilities in drug development, particularly in generating de novo, or from scratch, mini-proteins, peptides, and antibody inhibitors. The potential for successful applications of this technology is vast, including accelerated drug discovery, reduced costs, increased success rates, and highly targeted, tailored therapeutics with the possibility for fewer side effects. That being said, usage of AI in drug discovery is still in its infancy and faces challenges, including confronting the unpredictable nature of human biology and broader success accuracy. Still, such rapid development in little more than half a decade is impressive and justifies the excitement.
One area of AI application that has seen rising attention as of late is in the development of peptide therapeutics. In the past few years alone, several new AI models—AlphaFold, ProteinMPNN, BindCraft, Peptide2Mol, and many more—have been created that have streamlined a variety of steps in peptide drug discovery, including initial design, optimization, and even preliminary computational evaluation.
As compared to small molecule or mini-proteins, peptide drugs are considerably smaller, have high target specificity, and employ impressive safety profiles due to their ability to mimic natural body molecules. Some of the most promising treatments out of this class of drugs include Ozempic, Forteo, and, famously, insulin. One of the major drawbacks of peptide drugs, however, is their tendency for degradation by cellular enzymes, a negative byproduct of their innate similarities to regular cellular molecules. Furthermore, bioavailability, or a peptide’s ability to successfully reach its intended target in order to have an active effect on the body, is a consistent challenge due to difficulties in peptide absorption through mucosal membranes—oral or nasal pathways. With such intrinsic difficulties, working with peptides for therapeutic purposes has been historically tedious.
New techniques, however, are attempting to overcome this hurdle. Utilizing an emerging field known as peptidomimetics, which stands at the intersection of chemistry, pharmacology, and structural biology, researchers have developed ways to modify otherwise unusable peptides in the hopes of revitalizing their medicinal potential. Peptidomimetics modifications include altering portions or individual amino acids in a peptide chain to make it less likely to be degraded by cellular enzymes and increasing the peptide’s ability to reach and bind to it’s target.
Combined with AI applications, such as BindCraft, which could provide an initial design that researchers could then modify via peptidomimetics to be more therapeutically applicable, peptides as a class of drugs are showing promising pharmacological potential. In just a few years, several AI applications key for peptide development have been released to the general scientific community at an exponential rate. If the current challenges can be overcome, researchers stand in an incredible position to harness the incredible capabilities of trained AI models in generating faster, cheaper, and more accurate computational models in drug development. With the world moving ever closer towards an AI-integrated future and the field of peptidomimetics redefining peptide drug potential, scientists have an unprecedented opportunity to revolutionize peptide drug development, unlocking new avenues for therapeutics possibilities and ushering in a new era for pharmacology.
Recent Comments