Guiding a language-model based protein design method towards MHC Class-I immune-visibility targets in vaccines and therapeutics
Bibliographic record
Abstract
Proteins have an arsenal of medical applications that include disrupting protein interactions, acting as potent vaccines, and replacing genetically deficient proteins. While therapeutics must avoid triggering unwanted immune-responses, vaccines should support a robust immune-reaction targeting a broad range of pathogen variants. Therefore, computational methods modifying proteins' immunogenicity without disrupting function are needed. While many components of the immune-system can be involved in a reaction, we focus on Cytotoxic T-lymphocytes (CTLs). These target short peptides presented via the MHC Class I (MHC-I) pathway. To explore the limits of modifying the visibility of those peptides to CTLs within the distribution of naturally occurring sequences, we developed a novel machine learning technique, CAPE-XVAE. It combines a language model with reinforcement learning to modify a protein's immune-visibility. Our results show that CAPE-XVAE effectively modifies the visibility of the HIV Nef protein to CTLs. We contrast CAPE-XVAE to CAPE-Packer, a physics-based method we also developed. Compared to CAPE-Packer, the machine learning approach suggests sequences that draw upon local sequence similarities in the training set. This is beneficial for vaccine development, where the sequence should be representative of the real viral population. Additionally, the language model approach holds promise for preserving both known and unknown functional constraints, which is essential for the immune-modulation of therapeutic proteins. In contrast, CAPE-Packer, emphasizes preserving the protein's overall fold and can reach greater extremes of immune-visibility, but falls short of capturing the sequence diversity of viral variants available to learn from. Source code: https://github.com/hcgasser/CAPE (Branch: CAPE_1.1)
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 | 0.002 |
| 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 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
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".