ProteinVAE: Variational AutoEncoder for Translational Protein Design
Bibliographic record
Abstract
Abstract There have recently been rapid advances in deep learning models for protein design. To demonstrate proof-of-concept, these advancements have focused on small proteins with lots of data for training. This means that they are often not suitable for generating proteins with the most potential for high clinical impact –due to the additional challenges of sparse data and large size many therapeutically relevant proteins have. One major application that fits this category is gene therapy delivery. Viral vectors such as Adenoviruses and AAVs are a common delivery vehicle for gene therapy. However, environmental exposure means that most people exhibit potent pre-existing immune responses to many serotypes. This response, primarily driven by neutralizing antibodies, also precludes repeated administration with the same serotype. Rare serotypes, serotypes targeting other species, and capsid engineering, have all been deployed in the service of reducing neutralization by pre-existing antibodies. However, progress has been very limited using conventional methods and a new approach is urgently needed. To address this, we developed a variational autoencoder that can generate synthetic viral vector serotypes without epitopes for pre-existing neutralizing antibodies. A compact generative computational model was constructed, with only 12.4 million parameters that could be efficiently trained on the limited natural sequences (e.g., 711 natural Adenovirus hexon sequences with average length of 938 amino acids). In contrast to the current state-of-the-art, the model was able to generate high-quality Adenovirus hexon sequences that were folded with high confidence by Alphafold2 to produce structures essentially identical to natural hexon structures. Molecular dynamics simulations confirmed that the structures are stable and protein–protein interfaces are intact. Local secondary structure and local mobility is also comparable with natural serotype behavior. Our model could be used to generate a broad range of synthetic adenovirus serotype sequences without epitopes for pre-existing neutralizing antibodies in the human population. It could be used more broadly to generate different types of viral vector, and any large, therapeutically valuable proteins, where available data is sparse.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".