Regularized indirect learning improves phage display ligand discovery
Bibliographic record
Abstract
Phage display is commonly employed for the discovery of high affinity ligands to biomolecular targets. However, ranking the discovered ligands for their affinity and specificity to the target is obscured by genetic amplification bias and amplification of target-unrelated phage, resulting in inefficient experimental validation and potentially intractable discovery. Here, we describe the use of indirect machine learning (ML) to improve the efficient discovery of target-specific peptide ligands from next-generation sequencing (NGS) data. We combine peptide sequence information (input) with experimental fitness scores (output) of the individual peptide performance across the rounds of bio-panning in a bidirectional long short-term memory (BiLSTM) architecture. Because the fitness scores contain bias, we use regularization to facilitate limited indirect learning and effectively process the peptide sequence information, while still using the predicted fitness scores to rank the peptides. Peptides containing high-affinity binding motifs to our target were ranked by the regularized model more than threefold higher, compared to any combination of experimental fitness scores. Baseline models of random forest (RF) and -nearest neighbor (KNN) demonstrated slightly lower performance but also demonstrated the importance of regularization. However, the BiLSTM model emerged as the most robust, as it was less sensitive to the peptide representation and the specific fitness score used. Shapley residue analysis generated interpretable structure-activity-relationship (SAR) by providing insight into predicted affinity-driving residues and physicochemical properties across the entire peptide and as well as at motif-specific positions. We expect that this approach will elucidate high-affinity ligands against a multitude of targets, vastly improving the discovery capability of phage display.
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 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 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".