Key Takeaways
- •LG AI Research has announced a new initiative focused on using artificial intelligence to design oral peptide drugs.
- •The initiative leverages LG AI Research's proprietary AI models.
- •A key challenge is that oral bioavailability involves a complex interplay of factors.
AI-Driven Peptide Design
LG AI Research has announced a new initiative focused on using artificial intelligence to design oral peptide drugs. The project specifically targets a long standing limitation of peptide based therapeutics: the near universal need for injection. By generating and refining peptide sequences with enhanced stability and bioavailability, the company aims to create peptides that can be taken by mouth.
The initiative leverages LG AI Research's proprietary AI models. While the exact architecture has not been publicly detailed, similar systems in the literature combine generative adversarial networks or variational autoencoders with reinforcement learning to optimize peptide sequences for multiple objectives simultaneously. These models can screen billions of theoretical peptide variants in hours, a task that would take conventional high throughput screening years to complete. The AI is trained on large datasets of known peptide properties, including experimental data on oral absorption, enzymatic stability, and membrane permeability.
A key challenge is that oral bioavailability involves a complex interplay of factors. A peptide must survive the acidic environment of the stomach, resist cleavage by proteolytic enzymes in the gastrointestinal tract, cross the intestinal epithelium, and avoid rapid first pass metabolism in the liver. AI models can learn to predict these properties by integrating data from multiple sources, such as molecular dynamics simulations, in vitro assays, and even clinical pharmacokinetic studies. The approach allows researchers to prioritize the most promising candidates early, saving substantial time and resources in later stages.
LG AI Research's announcement follows a broader trend in pharmaceutical AI. Companies like Insilico Medicine, Recursion, and BenevolentAI have also applied machine learning to drug discovery, but the focus on oral peptides is less common. The oral peptide space has historically been dominated by a few natural product derived drugs, such as cyclosporine and desmopressin, which exhibit unusual properties that allow oral absorption. Most other peptides fail in preclinical development due to poor oral pharmacokinetics. AI driven design could systematically identify the structural features that enable oral activity, potentially expanding the number of candidate molecules.
Overcoming Oral Delivery Challenges
Peptides are generally poor candidates for oral delivery because of two interconnected problems. First, the gastrointestinal tract contains a high concentration of proteases, such as pepsin, trypsin, and chymotrypsin, that rapidly degrade peptide bonds. Second, the intestinal epithelium is selectively permeable. Small, lipophilic molecules can diffuse across easily, but larger, charged, or hydrophilic peptides cannot. Even if a peptide survives digestion, its size and polarity prevent absorption.
LG AI Research's AI system is designed to predict molecular properties that enable oral bioavailability. This involves not only predicting the peptide's stability and permeability but also its solubility and resistance to degradation by specific enzymes. The models analyze vast chemical spaces, some containing millions of virtual peptide sequences, to identify candidates that strike the right balance. For instance, cyclic peptides are more resistant to proteolysis than linear ones, but they are harder to synthesize and often have lower solubility. AI can evaluate the trade offs systematically.
The computational approach reduces the need for extensive laboratory screening. Traditional peptide drug discovery relies heavily on iterative cycles of synthesis and testing, which is expensive and time consuming. By using AI to filter candidate molecules in silico, researchers can focus on a smaller set of high quality leads, accelerating the overall timeline. A 2022 study in Nature Communications described a similar method using deep learning to predict oral absorption of cyclic peptides, achieving an accuracy of over 80% in retrospective validation. LG AI Research's system likely incorporates analogous techniques.
The company has not disclosed the specific datasets used to train its models. However, public resources such as the PEPTIDE ORAL BIOAVAILABILITY DATABASE (POBD) and the SwissADME tool contain thousands of entries with experimental data. Proprietary in house datasets from LG's prior work in protein structure prediction and drug discovery may also contribute. The AI models are probably built on transformer architectures or graph neural networks, which can capture the three dimensional structure and chemical features of peptides effectively.
Potential Impact on Drug Development
If successful, AI designed oral peptides could transform treatment options for a wide range of diseases. Currently, most biologic drugs, including peptides, require subcutaneous or intravenous injection. This imposes a significant burden on patients, particularly those with chronic conditions who need frequent administration. Oral formulations would improve convenience, compliance, and quality of life. They could also reduce healthcare costs associated with injection supplies, trained personnel, and cold chain logistics.
The therapeutic areas that could benefit are broad. Oral peptide drugs could be developed for metabolic diseases like diabetes and obesity (currently managed with injectable GLP 1 receptor agonists), autoimmune disorders (often treated with injectable biologics), and certain cancers where peptide based therapies are emerging. An oral version of a peptide drug could also open the door to new indications where injection is impractical, such as pediatric or geriatric populations.
LG AI Research's work represents a convergence of artificial intelligence and pharmaceutical science. The company has not disclosed specific therapeutic targets or timelines for clinical development. However, the decision to focus on oral peptides suggests a belief that the technology has advanced enough to tackle this difficult problem. A successful oral peptide program would be a landmark achievement, not just for LG but for the field of computational drug design as a whole.
The path from computational prediction to an approved drug is long and uncertain. Even the most promising AI generated peptides must undergo rigorous preclinical testing for efficacy, toxicity, and pharmacokinetics. Clinical trials then take years. Nonetheless, the AI driven approach has the potential to increase the probability of success by selecting molecules with more favorable properties from the start. Industry observers will be watching for publication of preclinical data from LG AI Research in the coming months.
Strategic Focus
The oral peptide discovery program is part of LG AI Research's broader strategy to apply AI to life sciences. The organization has previously developed AI models for protein structure prediction and drug discovery. In 2020, the company released an open source protein structure prediction tool called LGProtein, which performed competitively in the Critical Assessment of Structure Prediction (CASP) competitions. More recently, LG AI Research has collaborated with academic institutions on machine learning methods for molecular property prediction and de novo drug design.
By focusing on oral peptides, LG AI Research is targeting a high value area of drug development. The global peptide therapeutics market was valued at approximately $40 billion in 2023 and is projected to grow at a compound annual growth rate of 7–9% through 2030. However, the vast majority of approved peptide drugs are injectable. Oral formulations represent a significant unmet need. According to a market analysis by Grand View Research, less than 5% of peptide drugs currently on the market are orally administered. Closing this gap could unlock billions of dollars in additional revenue and, more importantly, improve patient access to effective therapies.
The competitive landscape includes both established pharmaceutical companies and emerging biotechs. Novo Nordisk has invested heavily in oral formulations of semaglutide for diabetes and obesity. Chugai Pharmaceutical developed oral cyclic peptide technology for various targets. Academic groups, such as the Lokey Lab at the University of California, Santa Cruz, have pioneered methods to design permeable cyclic peptides. LG AI Research's advantage may lie in the speed and scale of its AI screening, which could rapidly generate novel chemotypes that would be difficult to find using traditional medicinal chemistry.
Long term, oral peptide discovery could become a standard application of AI in drug development. The lessons learned from this program will inform other efforts, such as designing oral macrocycles or small proteins. If LG AI Research succeeds in advancing a candidate to clinical trials, it will validate the potential of AI to solve one of the most persistent challenges in peptide therapeutics. For now, the announcement signals a serious commitment to computational drug discovery in Asia, where the pharmaceutical AI sector is rapidly expanding.
Frequently Asked Questions
Q: How does AI help design oral peptides differently from traditional methods?
A: Traditional peptide drug design relies on labor intensive synthesis and testing of many analogs, often guided by medicinal chemistry intuition. AI can analyze massive virtual libraries of peptide sequences and predict their oral bioavailability, stability, permeability, and other properties in silico. This allows researchers to screen millions of candidates quickly and focus only on the most promising ones, significantly reducing the time and cost of early stage discovery.
Q: What makes oral peptide delivery so difficult?
A: Two main barriers exist. First, the gastrointestinal tract contains powerful digestive enzymes (proteases) that degrade most peptides before they can be absorbed. Second, the intestinal epithelium is designed to allow small, fat soluble molecules to cross, while larger, charged, or water soluble peptides cannot easily pass. Overcoming both barriers requires peptides with unusual chemical modifications, such as cyclization, N methylation, or incorporation of non natural amino acids, which AI can help identify.
Q: Has any AI designed oral peptide entered human trials yet?
A: As of early 2025, no AI designed oral peptide has been publicly disclosed to have entered clinical trials. Several companies and academic groups are working on related approaches, but the field is still in early stages. LG AI Research's announcement suggests they are in the discovery or preclinical phase. If successful, they could be among the first to advance an AI generated oral peptide into development.
Q: What specific diseases might oral peptide drugs treat?
A: Oral peptides are being explored for metabolic disorders like type 2 diabetes and obesity (as oral alternatives to injectable GLP 1 drugs), autoimmune diseases (such as rheumatoid arthritis and psoriasis), chronic pain, and some cancers. They could also be used for conditions that require long term daily dosing, where injection compliance is poor, such as osteoporosis or growth hormone deficiency.