Biology subjects

Bikias, T.

Publications and source records attributed to Bikias, T..

6 recordsLinked to original sources

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↗

bioRxiv2025

Generative design of antibody Fc-variants with synthetic and programmable functional profiles

Beyond antigen recognition, antibodies direct diverse immune effector functions through their constant (Fc) domain. While the Fc domain is central to antibody biology and therapeutic efficacy, our understanding of how Fc sequence encodes function remains limited, as most of Fc sequence space has not been experimentally mapped or linked to Fc-receptor engagement. Furthermore, the extensive overlap in Fc-receptor binding sites on the Fc domain has impeded efforts to engineer antibodies with tailored, multi-receptor engagement profiles that can precisely control downstream immunity. Here we introduce a novel framework for Fc engineering that integrates protein engineering with deep learning to rationally predict and engineer antibody Fc function. Using a yeast-based, aglycosylated Fc display system, we performed deep mutational scanning across the entire human IgG1 Fc domain, allowing the rational design of a diverse combinatorial library of more than 108 Fc-variants. This library was sorted based on binding to a panel of eight canonical Fc-receptors, and the resulting populations were deep sequenced to generate a high-quality dataset comprising millions of unique Fc sequences annotated with their respective Fc-receptor binding profiles. Deep learning-based classifiers trained on this dataset accurately predicted Fc-receptor binding activity from Fc sequence across all Fc-receptors tested. We further developed FcGPT, a domain-specific autoregressive protein language model pre-trained on over three million unique Fc sequences, and refined by post-training through reinforcement learning with experimental feedback (RLXF) and synthetic verifiers. FcGPT enables the computational design of novel Fc-variants with user-defined Fc-receptor binding profiles, providing a foundational tool for understanding and programming antibody-mediated immunity.

bioengineering↗

bioRxiv2025

Dissecting serum polyclonal antibody escape to SARS-CoV-2 variants by deep mutational learning

The rapid emergence of SARS-CoV-2 variants harboring multiple receptor-binding domain (RBD) mutations continues to challenge the efficacy of vaccines and antibody therapeutics. While deep mutational scanning (DMS) has been instrumental in mapping single-mutation effects on antibody binding and immune escape, it remains limited in its ability to assess combinatorial mutational landscapes. Here, we extend deep mutational learning (DML), a method integrating combinatorial mutagenesis, yeast surface display, deep sequencing, and machine learning, to analyze serum polyclonal antibody escape. Human sera from COVID-19-vaccinated individuals were screened against diverse RBD variant libraries, validating 300 serum-variant interactions across 10 individuals by comparing predicted and observed binding to 30 RBD variants, confirming accurate mapping of binding and escape profiles. Model performance remained consistent across machine learning architectures, suggesting that serum binding and escape are governed by distinct, localized RBD sequence features. Notably, escape profiles were highly individualized, whereas binding signatures were more conserved, reflecting convergent epitope targeting. This approach highlights the potential of DML to generalize beyond observed RBD variants, assess cohort-specific immune breadth, and inform vaccine and therapeutic design in the face of viral evolution.

immunology↗

bioRxiv2025

PLMFit : Benchmarking Transfer Learning with Protein Language Models for Protein Engineering

Protein language models (PLMs) have emerged as a useful resource for protein engineering applications. Transfer learning (TL) leverages pre-trained parameters to extract features to train machine learning models or adjust the weights of PLMs for novel tasks via fine-tuning through back-propagation. TL methods have shown potential for enhancing protein predictions performance when paired with PLMs, however there is a notable lack of comparative analyses that benchmark TL methods applied to state-of-the-art PLMs, identify optimal strategies for transferring knowledge and determine the most suitable approach for specific tasks. Here, we report PLMFit, a benchmarking study that combines, three state-of-the-art PLMs (ESM2, ProGen2, ProteinBert), with three TL methods (feature extraction, low-rank adaptation, bottleneck adapters) for five protein engineering datasets. We conducted over >3,150 in silico experiments, altering PLM sizes and layers, TL hyperparameters and different training procedures. Our experiments reveal three key findings: (i) utilizing a partial fraction of PLM for TL does not detrimentally impact performance, (ii) the choice between feature extraction and fine-tuning is primarily dictated by the amount and diversity of data and (iii) fine-tuning is most effective when generalization is necessary and only limited data is available. We provide PLMFit as an open-source software package, serving as a valuable resource for the scientific community to facilitate the feature extraction and fine-tuning of PLMs for various applications. ONE SENTENCE SUMMARYPLMFit is a comparative analysis aimed at identifying the most effective strategies for transfer knowledge from protein language models by benchmarking fine-tuning techniques on a range of protein engineering tasks.

bioinformatics↗

bioRxiv2024

Synthetic coevolution reveals adaptive mutational trajectories of neutralizing antibodies and SARS-CoV-2

The Covid-19 pandemic showcases a coevolutionary race between the human immune system and SARS-CoV-2, mirroring the Red Queen hypothesis of evolutionary biology. The immune system generates neutralizing antibodies targeting the SARS-CoV-2 spike proteins receptor binding domain (RBD), crucial for host cell invasion, while the virus evolves to evade antibody recognition. Here, we establish a synthetic coevolution system combining high-throughput screening of antibody and RBD variant libraries with protein mutagenesis, surface display, and deep sequencing. Additionally, we train a protein language machine learning model that predicts antibody escape to RBD variants. Synthetic coevolution reveals antagonistic and compensatory mutational trajectories of neutralizing antibodies and SARS-CoV-2 variants, enhancing the understanding of this evolutionary conflict.

synthetic biology↗