Predicting Peptide Aggregation with Protein Language Model Embeddings.
Peptide aggregation prediction just got a major boost, thanks to deep learning and protein language models. Researchers at Novo Nordisk rolled out PALM, a tool that uses embeddings from massive protein language models to forecast which peptides will clump together—an issue that can wreck peptide-based therapeutics and complicate everything from manufacturing to disease research.
J Chem Inf Model
by Eschbach E, Deibler K, Korani D et al.
“Predicting Peptide Aggregation with Protein Language Model Embeddings. Eschbach E(1), Deibler K(1), Korani D(1), Swanson S(1). Author information: (1)Molecular AI, Novo Nordisk, Lexington, Massachusetts 02421, United States. Amyloid fibrils, a form of peptide aggregates, are associated with multiple diseases and hinder the development of therapeutics. The experimental characterization of aggregating peptides is resource-intensive, and data are scarce, limiting the development of accurate models. We present a deep-learning model, PALM (Predicting Aggregation with Language Model embeddings), which uses transfer learning to predict aggregation from embeddings extracted from a pretrained protein language model (pLM). PALM is trained on the WaltzDB-2.0 dataset to classify peptides and identify aggregation-prone regions within a sequence at single-residue resolution. Compared to existing models, it exhibits competitive performance on diverse held-out experimental datasets. We find that PALM fails to identify single mutations that increase the rate of aggregation of the amyloid beta peptide; however, training the PALM architecture on a larger dataset, CANYA NNK1-3, substantially improves performance in this task. These results show that transfer learning with pLM embeddings improves performance when training on small datasets, but highlight that challenging tasks, such as predicting the effect of single mutations, require more experimental data.”
Here’s the problem: Amyloid fibrils, a notorious kind of peptide aggregate, pop up in a ton of diseases and make it tough to develop new compounds. Running lab experiments to test aggregation is slow and expensive. Data is thin, so machine learning models often stall. PALM changes the game by leaning on transfer learning. Instead of starting from scratch, it taps into protein language models already trained on huge datasets. It then fine-tunes these models using the WaltzDB-2.0 dataset, homing in on aggregation-prone regions down to individual amino acids.
Key findings:
PALM holds its own against other aggregation predictors on tough, real-world datasets.
The model struggled to catch single-point mutations in amyloid beta that ramp up aggregation—until the team trained it on a much bigger dataset, CANYA NNK1-3. With more data, performance shot up.
The takeaway: Transfer learning from pretrained protein models is a win for peptide aggregation prediction, especially when datasets are small. But if you want to nail down subtle effects, like single-residue changes, you still need lots of experimental data.
For anyone deep into peptide design or troubleshooting unexpected aggregation, these results are a clear signal: smarter algorithms can make your life easier, but high-quality data is still king. Check out the peptide research index for more on the latest tools and breakthroughs. The future of peptide aggregation prediction looks a lot less like guesswork.
Related Reading
The STRIDE Trial and Semaglutide: Implications for Clinical Vascular Practice.
News · J Med ChemStructure-Based Adaptation of a SARS-CoV-2 Neutralizing Peptide to New Virus Variants.
News · J Nucl MedCombining an α(v)β(6)-Targeted (177)Lu-Based Peptide Receptor Radionuclide Therapy with Olaparib to Boost Therapeutic Efficacy in Pancreatic Cancer.
For Research Use Only
All content published on Pushing Peptides is intended for educational and informational purposes only. The information provided is not intended as medical advice, diagnosis, or treatment. Peptides discussed in this article are research compounds and are not approved for human therapeutic use by the FDA or any other regulatory agency. All studies referenced involve animal models or in vitro research unless otherwise stated. Consult a qualified healthcare professional before making any decisions related to your health. Pushing Peptides does not sell peptides — we are a vendor directory and educational resource.