AssayRouter: Historical Utility Priors for Frozen Molecular Predictor Routing

2026-09-29

AI Summary

This study addresses the challenge of efficiently selecting an optimal source model from a frozen model zoo for novel molecule detection, where labels are scarce and predictor parameters remain inaccessible. The proposed method constructs a candidate utility prior using historical detection data, pioneering the treatment of completed detections as pseudo-targets to quantify candidate utility. Furthermore, a shared regressor predicts the loss reduction upon introducing local models, enabling identity-free dynamic routing and convex combination optimization. Experimental results demonstrate that this approach significantly reduces negative log-likelihood in external regression detection tasks and validates mapping generalization across interface families. By overcoming traditional reliance on source-specific information, this work confirms the transferability of historical supervisory signals under frozen interfaces.

Abstract

Laboratories often face a new molecular assay with 16-64 labels and a bank of predictors whose training data and parameters are unavailable. The practical question is which frozen outputs to include in a small local model. AssayRouter treats completed assays as pseudo-targets and labels each candidate by its post-fit utility: the reduction in held-out discovery loss when the candidate is added to the local target predictor. A shared regressor learns to predict this utility from candidate behavior on the support set, without source identity; on a new assay, one frozen ranking selects four sources and separate labels fit a convex combiner. We train only on completed ChEMBL-MT assays and evaluate 24 external regression assays across six frozen interface families. AssayRouter-C lowers strict four-call negative log-likelihood (NLL) by 0.0409 relative to Support-CV@4. Frozen candidate-label permutations confirm that candidate-utility correspondence carries the transferred information, and leave-one-interface-out training shows that the mapping generalizes to unseen predictor families. Completed assays therefore provide transferable supervision for scarce-label routing through frozen prediction interfaces.

Problem

• frozen molecular predictor routing
• molecular assay
• model selection
• few-label learning
• predictor utility

Innovation

• Frozen Predictor Routing
• Historical Utility Priors
• Molecular Assay
• Convex Combiner
• Transferable Supervision