Schifferli, T. | bioRxiv Science

bioRxiv2026

CAPO: Compute-Aware Automated Protein-Model Optimization

Protein models such as Evolutionary Scale Modeling (ESM), AlphaFold and Boltz have transformed computational protein science, yet a pretrained model rarely transfers to a particular laboratory's assay without adaptation to its data and experimental setting, and that adaptation is technically demanding and spends real compute budget. Existing agentic systems can automate model training, yet typically treat fine-tuning as a reliable tool call: feasibility and cost are estimated rather than measured, and infeasibility is discovered only after compute has been spent. We formulate protein-model adaptation as a gated decision problem: given a task, raw data and a compute budget c_max, a system should decide whether and how to adapt a model before committing expensive computation. We introduce CAPO (Compute-Aware Automated Protein-Model Optimization), a multi-agent system built around one rule: gate before you train. CAPO measures feasibility on the target accelerator, projects cost against the available budget, runs a canary inside the actual training loop and maintains an auditable artifact trail; critically, it can decline an unaffordable adaptation and select a cheaper alternative. We evaluate CAPO, without system changes, on three protein-modeling tasks: (i) SARS-CoV-2 host-range prediction, (ii) hit-to-lead affinity ranking and (iii) temporal variant-effect reclassification. On host range, CAPO partially fine-tuned an 8-million-parameter ESM-2 model to a species-averaged MCC of 0.826, compared with 0.743 for a hand-designed expert transformer; it matched a general coding agent while using 57.4% less GPU time and reduced failed training trials from four to zero. For hit-to-lead affinity ranking, CAPO rejected Boltz-2 fine-tuning and instead trained an assay-calibrated lightweight head over frozen representations, improving ranking on 11 of 15 held-out targets in 1.68 GPU-hours. On temporally held-out ClinVar variants, rank-4 low-rank adaptation of ESM-2 reached an MCC of 0.478, compared with 0.252 for a one-hot baseline and 0.000 for the frozen zero-shot score. These results come from single runs, without a gate ablation and with one generalist-agent comparison. Within those limits, they show that measured admission control can make protein-model adaptation not only automated but selective: deciding what to compute, what not to compute and why.

bioinformatics↗

bioRxiv2026

Predicting viral host range for reverse zoonosis surveillance

Zoonotic transmission of viruses has triggered major outbreaks throughout history. However, human-to-animal transmission, or reverse zoonotic transmission, also represents an underappreciated public health threat. Reverse zoonoses can establish novel animal reservoirs that fuel viral evolution and enable eventual spillback into humans, underscoring the importance of defining viral host range. Current approaches for host range prediction are limited. Experimental methods are retrospective and rely on brute-force testing in a small number of models, whereas computational approaches often lack experimental grounding. Here, we present a platform that integrates high-throughput experimental screening with deep learning to predict viral host range from sequence. We screened a yeast-displayed SARS-CoV-2 receptor-binding domain library against 15 ACE2 orthologs from diverse species, and deep sequenced the ACE2-binding and non-binding populations to generate an atlas of 8.8 million unique RBD:ACE2 interaction measurements. These data were used to train a transformer-based deep learning model that predicts RBD:ACE2 compatibility across animal species from sequence alone, and generalizes in a zero-shot manner to ACE2 orthologs not observed during training. Critically, the model pairs each prediction with a distance-based uncertainty estimate, allowing hundreds of candidate species to be triaged by both predicted host compatibility and confidence. This framework enables the rapid prioritization of animal species for SARS-CoV-2 surveillance and provides a general strategy for anticipating viral host range.

bioengineering↗