Biology subjects

Sormanni, P.

Publications and source records attributed to Sormanni, P..

At least 19 recordsLinked to original sources

bioRxiv2026

Engineering armoured in vivo CAR T cells through targeted delivery and transient mRNA gating

In vivo generation of chimeric antigen receptor (CAR) T cells offers an off-the-shelf, scalable alternative to ex-vivo therapies, but is constrained by inefficiency, toxicity and durability. We report the first armoured in-vivo CAR T platform with tunable, transient activity, that enhances safety and efficacy. Our platform utilises CD8 targeted lipid nanoparticles (LNP) to transiently deliver CAR and armouring membrane-bound IL12 mRNAs. Payloads utilize a T-cell-restricted (T-trex) strategy to constrain CAR and IL-12 expression to T cells, enabling potent antigen-dependent activity with spatio-temporal control. Selective targeting of T-cells with a T-trex CAR showed enhanced expression. The addition of T-trex IL12 mRNA provides a tunable, localised amplification of CAR function, enhancing antigen-dependent cytotoxicity and cytokine production with robust tumour control. These findings establish a novel dose-efficient armoured in-vivo CAR T approach and general framework for programmable immune cell engineering.

immunology↗

bioRxiv2026

Structural Plausibility Without Binding Specificity: Limits of AI-Based Antibody-Antigen Structure Prediction Confidence Scores

Antibody-antigen binding prediction remains a central challenge for AI-driven therapeutic discovery, particularly in discriminating cognate interactions from structurally plausible but incorrect pairings. We present a controlled, AI-method- and antibody-format-agnostic evaluation framework that measures binding specificity under realistic conditions. Using 106 experimentally determined single-chain antibody (nanobody)-antigen complexes and 11,342 shuffled non-cognate pairings, we benchmarked publicly-available state-of-the-art structure prediction methods (AlphaFold3, Boltz-2, Chai-1). Although the methods tested often generated geometrically plausible complexes, internal confidence metrics (ipTM) frequently failed to discriminate correct from incorrect pairings. Increased sampling improved structural refinement but not pairing discrimination, indicating that computational resources are better allocated across independent seeds and explicit negative controls. We conclude that internal confidence scores are not inherently calibrated to binding specificity and require validation against realistic decoys. To enable community benchmarking and method development, we release [~]1.8 million AI-generated complex structures and guidance for the benchmarks ahead.

bioinformatics↗

bioRxiv2026

Disulphide and sequence-encoded conformational priors guide nanobody structure prediction

Nanobody binding is largely governed by the HCDR3 loop, which adopts distinct placement regimes relative to the framework: compact, framework-contacting (kinked blueprint) and solvent-exposed (extended blueprint). Many nanobodies also contain additional cysteines that form non-canonical disulphide bonds, imposing covalent constraints on binding-loop conformations. Current structure predictors are typically trained and benchmarked with smooth coordinate-based objectives, so models may appear reasonable under root-mean-square deviation (RMSD), while adopting an incorrect HCDR3 blueprint or failing to recover the native disulphide connectivity, impacting paratope geometry and functional interpretation. Here, we show that the HCDR3 blueprint is predictable from sequence alone, allowing for explicit constraints during modelling. We implement these principles into NbForge, a lightweight nanobody folding model that incorporates blueprint- and disulphide-aware inductive biases and is trained with filtered self-distillation. NbForge improves recovery of HCDR3 blueprint and non-canonical disulphide formation over previous lightweight models and achieves coordinate accuracy at par to state-of-the-art, large, resource-intensive predictors, while running at sub-second inference speed. We show that using NbForge monomer models as templates further improves the success rate of predicting nanobody-antigen complexes. Together, these results motivate blueprint- and disulphide-aware benchmarks for nanobody modelling beyond RMSD, and show that appropriate inductive biases can close the performance gap to heavyweight predictors. We make the sequence classifier (NbFrame) and NbForge available for download and via a user-friendly web server.

bioinformatics↗

bioRxiv2026

Molecularly specific solubilization of therapeutic antibodies

Understanding the effects of formulation excipients on protein solubility is a key part of physical chemistry and pharmaceutical sciences. While excipients are routinely employed to reduce the self-association of biologic drugs, their mechanisms of action remain poorly understood and are often assumed to be broadly nonspecific. Using a high-throughput combinatorial droplet microfluidic platform, we systematically survey and quantify how common pharmaceutical excipients affect the solubility of a diverse panel of therapeutic monoclonal antibodies (mAbs). We show that, while excipients are generally solubilizing, their effects vary substantially across different mAbs, with excipient-specific solubilization scores spanning dynamic ranges of approximately 7-fold to >200-fold across the antibody panel. Histidine, arginine and sodium chloride, in particular, engage in interactions characterized by unique molecular specificity, whereas sucrose effects are largely governed by nonspecific, solvent-mediated interactions. Correlating excipient performance with dynamical mAb molecular features from solvated full-length homology models allows us to dissect and quantify the relative contributions of molecular features governing excipient-mediated solubilization. We envision this new physicochemical understanding lays the groundwork for rational excipient selection and bespoke formulation design, with direct implications for accelerating protein therapeutic development for preclinical scenarios.

biophysics↗

bioRxiv2025

SurfDiff: protein surface profiling for selective or broadly reactive epitope prioritisation in binder and immunogen design

Rational selection of protein-surface patches - epitopes or ligandable sites - is essential for developing targeted antibodies, binding proteins, peptides, or small-molecule ligands, and for informing immunogen design in vaccines. We present SurfDiff, a structure- and sequence-informed framework for protein surface comparison. SurfDiff supports one-to-one and one-to-many comparisons by combining local structural alignments with physicochemically and spatially aware neighbourhood analysis. It assigns residue-level uniqueness and similarity scores, which can be aggregated into surface-level similarity, selectivity, and a discriminability score that captures conservation across desired targets while penalising similarity to undesired off-targets. These scores, calculated solely from the targets without any knowledge of the binders, reliably predict known experimental selectivity for diverse binders including small molecules, peptides, antibodies and antibody mimetics across viral antigens, cytokine isoforms, GPCR subtypes and serum albumin. They also correlate strongly with binding and neutralisation data across pathogen variants, and substantially outperform commonly used bioinformatic metrics such as sequence substitution matrices and structural similarity measures. SurfDiff provides a generalisable and interpretable approach to protein surface profiling, enabling selective or cross-reactive binder design, cross-species prioritisation, and facilitating immunogen selection. We make SurfDiff available as downloadable open-source software: (https://gitlab.developers.cam.ac.uk/ch/sormanni/surfdiff) and as a webserver: www-cohsoftware.ch.cam.ac.uk/index.php/surfdiff.

bioinformatics↗

bioRxiv2025

Improving nanobody structure prediction with self-distillation

Nanobodies are increasingly attractive therapeutic and biotechnological molecules, yet accurate structure prediction of their highly variable H-CDR3 loops remains a central challenge for machine learning models. Here, we investigate whether nanobody-specific structure prediction can be improved through curated synthetic data strategies. We systematically evaluate different data augmentation regimes, including self-distillation from unlabelled VHH sequences. To ensure structural plausibility of synthetic training samples, we develop NanoKink, the first sequence-based classifier of kinked versus extended H-CDR3 conformations, and apply stringent filtering criteria for non-canonical disulfide bond placement and confor-mational accuracy. On a curated benchmark enriched for challenging nanobody features, we show that, for a fixed training compute budget, a nanobody-specific model trained with filtered synthetic data significantly improves over baseline models and NanobodyBuilder2, achieving lower mean H-CDR3 RMSD and fewer structural violations, while remaining competitive with AlphaFold3 at approximately two orders of magnitude lower per-structure inference time. Our results highlight promising directions in synthetic data generation for nanobody structure modelling and provide a practical framework for optimisation of VHH structure prediction models.

bioengineering↗

bioRxiv2025

Deep learning assessment of nativeness and pairing likelihood for antibody and nanobody design with AbNatiV2

1Immune systems create antibodies that balance good binding and stability with low toxicity and self- reactivity. Quantifying the nativeness of a candidate sequence - its likelihood of belonging to natural immune repertoires - has thus emerged as a valuable strategy for hit selection from synthetic libraries, optimisation and humanisation, and for guiding de novo design towards developable candidates. We previously introduced AbNatiV, a transformer-based VQ-VAE for nativeness assessment, which proved effective across multiple nanobody engineering tasks. However, AbNatiV1 operated on unpaired sequences, limiting applicability to conventional VH-VL antibodies. Moreover, its performance on nanobody nativeness was constrained by the limited number and diversity of nanobody repertoires available at the time. Here, we sequenced new camelid repertoires, curated additional recent datasets, and present AbNatiV2: an enhanced architecture comprising various models each trained on [≥] 20 million sequences. AbNatiV2 improves nanobody nativeness classification across held-out and diverse test sets, and more robustly detects nativeness changes upon CDR grafting. We also introduce p-AbNatiV2, a cross-attention model fine-tuned on 3.7 million paired human sequences. p-AbNatiV2 provides residue- and sequence-level humanness for VH/VL pairs and learns pairing-likelihood via noise-contrastive training. On held-out tests, it assigns the native pair a higher score in 74% of cases, substantially outperforming recent pairing models. Together, AbNatiV2 and p-AbNatiV2 extend nativeness assessment and engineering to both nanobodies and conventional antibodies, supporting design decisions at single-residue, Fv-sequence, and paired-domain levels. We make AbNatiV2 available as downloadable software and webserver.

bioengineering↗

bioRxiv2025

BLOSUM Is All You Learn - Generative Antibody Models Reflect Evolutionary Priors

Generative models have emerged as powerful tools for antibody sequence design, with recent studies demonstrating that log-likelihood scores from these models can correlate with binding affinity and potentially serve as effective ranking metrics. This raises a fundamental question: why should log-likelihood scores from generative models correlate with binding affinity? In this work, we investigate the biochemical basis of these model-derived log-likelihoods by comparing them with classical evolutionary similarity metrics. We find that BLOSUM similarity scores between designed and parental antibody sequences correlate strongly with measured binding affinity--on par with the predictive performance of a state-of-the-art diffusion-based generative model. Moreover, these BLOSUM scores also align closely with log-likelihoods from multiple generative models, suggesting that such models may be implicitly learning evolutionary priors encoded in substitution matrices. When computed with respect to a known binder, both BLO-SUM scores and log-likelihoods act as approximate measures of sequence distance from that reference. As this distance increases, the likelihood of a candidate being a binder decreases, explaining the observed correlation between these scores and binding affinity. In contrast, similarity scores based on position weight matrices (PWMs) and position-specific scoring matrices (PSSMs), which do not rely on knowledge of the parental sequence, show weaker and less consistent alignment with binding affinity, with performance depending on the background sequence data. Additionally, using consensus sequences in place of parental sequences to compute BLOSUM scores largely eliminates the observed correlation with affinity, underscoring the context-specific nature of the correlations. These findings highlight the potential of interpretable, evolution-inspired metrics to complement generative modeling in anti-body design, offering insights into both model behavior and biological relevance.

bioengineering↗

bioRxiv2025

Predicting Experimental Success in De Novo Binder Design: A Meta-Analysis of 3,766 Experimentally Characterised Binders

Designing high-affinity de novo protein binders has become increasingly tractable, yet in vitro prioritisation continues to depend on heuristics in the absence of systematic analysis. Here, we present a large-scale meta-analysis of 3,766 experimentally tested binders across 15 structurally diverse targets. Using a unified, high-throughput pipeline that re-predicts each binder-target complex with AF2 (initial guess and ColabFold), AF3 and Boltz-1, we extract over 200 structural, energetic and confidence features per design. We show that interface-focused metrics, most notably the AF3-derived interaction prediction Score from Aligned Errors (ipSAE) outperform commonly used scores such as ipAE and ipTM, with a significant 1.4-fold increase in average precision compared to ipAE. We further show that combining these metrics with orthogonal physicochemical interface descriptors, including Rosetta {Delta}G/{Delta}SASA and interface shape complementarity, improves predictive performance. While overall per-formance varies by target, simple linear models trained on a small number of AF3-derived features generalize well across datasets. We propose interpretable, target-agnostic filtering strategies, such as combining AF3 ipSAE_min rankings with structural filters, to improve precision in selecting binders for testing. Finally, we release the complete dataset establishing a community resource to benchmark and accelerate de novo binder discovery.

bioinformatics↗

bioRxiv2025

Prediction of Antibody Non-Specificity using Protein Language Models and Biophysical Parameters

The development of therapeutic antibodies requires optimizing target binding affinity and pharmacodynamics, while ensuring high developability potential, including minimizing non-specific binding. In this study, we address this problem by predicting antibody non-specificity by two complementary approaches: (i) antibody sequence embeddings by protein language models (PLMs), and (ii) a comprehensive set of sequence-based biophysical descriptors. These models were trained on human and mouse antibody data from Boughter et al. (2020) and tested on three public datasets: Jain et al. (2017), Shehata et al. (2019) and Harvey et al. (2022). We show that non-specificity is best predicted from the heavy variable domain and heavy-chain complementary variable regions (CDRs). The top performing PLM, a heavy variable domain-based ESM 1v LogisticReg model, resulted in 10-fold cross-validation accuracy of up to 71%. Our biophysical descriptor-based analysis identified the isoelectric point as a key driver of non-specificity. Our findings underscore the importance of biophysical properties in predicting antibody non-specificity and highlight the potential of protein language models for the development of antibody-based therapeutics. To illustrate the use of our approach in the development of lead candidates with high developability potential, we show that it can be extended to therapeutic antibodies and nanobodies.

bioinformatics↗

bioRxiv2025

NewroBus for the brain: humanized TfR1-targeting nanobodies with high BBB permeability and cargo transport capacity

Effective delivery of therapeutics to the brain is restricted by the blood-brain barrier (BBB). A strategy to overcome this limitation involves taking advantage of receptor-mediated transcytosis pathways, such as those mediated by transferrin receptor 1 (TfR1), which is highly expressed on brain endothelial cells and naturally transports iron-bound transferrin across the BBB. To exploit this mechanism, we immunized camelids with human TfR1 and cloned 470 VHH nanobody sequences from their B cells. From this repertoire, 24 nanobodies (TfR1b-Nbs) were identified that bind human TfR1 on the cell membrane. These nanobodies were screened for binding to human TfR1, lack of interference with transferrin binding and TfR1-mediated iron uptake, and the ability to cross the BBB via human TfR1-mediated transcytosis in newly generated humanized Tfr1h knock-in rats. To improve developability and reduce potential immunogenicity, selected TfR1b-Nbs were humanized and optimized with computational and artificial intelligence (AI) algorithms, enhancing humanness, solubility, and VHH-nativeness. Eight optimized TfR1b-Nbs retained BBB permeability and were fused to humanized anti-TNF nanobody inhibitors (TNFI- or TNFI-{beta}), generating 16 heterodimers. Fusion to these TNFIs served as a functional readout, confirming that TfR1b-Nbs can shuttle biologically active, BBB-impermeable payloads into the central nervous system (CNS). All heterodimers demonstrated CNS delivery after intravenous administration, and selected constructs also reached the brain via subcutaneous injection, maintaining high serum and cerebrospinal fluid (CSF) levels for up to 72 hours. A pilot study with one heterodimer showed that chronic administration in rats humanized for both transferrin and TfR1 caused no hematological toxicity or signs of anemia - a key safety concern when targeting TfR1. These results establish humanized TfR1b-Nbs - designated NewroBus - as promising BBB shuttles for the safe and effective therapeutic delivery of biologics to the brain.

bioengineering↗

bioRxiv2025

Two sites in the C-terminal β-chain tail mediate interactions of the chaperone clusterin with amyloid beta and other misfolded client proteins

Clusterin (CLU) is a constitutively secreted mammalian chaperone that binds in extracellular body fluids to misfolded client proteins to neutralise their toxicity and mediate their safe disposal by cell uptake and intracellular degradation. However, the regions of CLU critical for its interactions with misfolded proteins remain still largely unknown. To identify binding sites, we expressed a panel of CLU deletion and alanine-stretch mutants in a recently developed mammalian expression system. Mutant CLU molecules lacking detectable structural aberrations were subjected to functional analyses to compare their abilities with that of wild type CLU to bind to misfolded proteins and to inhibit protein aggregation. These analyses implicated two regions in the flexible {beta}-chain C-terminal tail of CLU as being important in the interactions of the chaperone with misfolded proteins, including aggregating the Alzheimers amyloid {beta}-peptide (A{beta}). We then designed in silico sequence-specific single-domain camelid nanobodies to confirm the function of the two putative client protein binding sites. Based on our experimental results and in silico binding site predictions, we suggest that the potent ability of CLU to promiscuously interact with many different misfolded proteins, regardless of their size or structure, arises from the location of multiple client protein binding sites in its flexible tail region.

biochemistry↗

bioRxiv2025

The SpyBLI cell-free pipeline for the rapid quantification of binding kinetics from crude samples

Accurate measurements of binding kinetics, encompassing equilibrium dissociation constant (KD), association rate (kon), and dissociation rate (ko,), are critical for the development and optimisation of high-affinity binding proteins. However, such measurements require highly purified material and precise ligand immobilisation, limiting the number of binders that can be characterised within a reasonable timescale and budget. Here, we present the SpyBLI method, a rapid and cost-effective biolayer interferometry (BLI) pipeline that leverages the SpyCatcher003-SpyTag003 covalent association, eliminating the need for both binder purification and concentration determination. This approach allows for accurate binding-kinetic measurements to be performed directly from crude mammalian-cell supernatants or cell-free expression mixtures. We also introduce a linear gene fragment design that enables reliable expression in cell-free systems without any PCR or cloning steps, allowing binding kinetics data to be collected in under 24 hours from receiving inexpensive DNA fragments, with minimal hands-on time. We demonstrate the methods broad applicability using a range of nanobodies and single-chain antibody variable fragments (scFvs), with affinity values spanning six orders of magnitude. By minimising sample preparation and employing highly controlled, ordered sensor immobilisation, our workflow delivers reliable kinetic measurements from crude mixtures without sacrificing precision. We expect that the opportunity to carry out rapid and accurate binding measurements in good throughput should prove especially valuable for binder engineering, the screening of next-generation sequencing-derived libraries, and computational protein design, where large numbers of potential binders for the same target must be rapidly and accurately characterised to enable iterative refinement and candidate selection.

biophysics↗

bioRxiv2025

Development of potent humanized TNFα inhibitory nanobodies for therapeutic applications in TNFα-mediated diseases

Tumor necrosis factor-alpha (TNF) is a key pro-inflammatory cytokine implicated in the pathogenesis of numerous inflammatory and autoimmune diseases, including rheumatoid arthritis, inflammatory bowel disease, and neurodegenerative disorders such as Alzheimers Disease. Effective inhibition of TNF is essential for mitigating disease progression and improving patient outcomes. In this study, we present the development and comprehensive characterization of potent humanized TNF inhibitory nanobodies (TNFINbs) derived from camelid single-domain antibodies. In silico analysis of the original camelid nanobodies revealed low immunogenicity, which was further reduced through machine-learning-guided humanization and developability optimization. The two humanized TNFI-Nb variants we developed demonstrated exceptional anti-TNF activity, achieving IC50 values in the picomolar range. Binding assays confirmed their high affinity for TNF, underscoring robust neutralization capabilities. These TNFI-Nbs present valid alternatives to conventional monoclonal antibodies currently used in human therapy, offering potential advantages in potency, specificity, and reduced immunogenicity. Our findings establish a solid foundation for further preclinical development and clinical translation of TNF-targeted nanobody therapies in TNF-mediated diseases.

biochemistry↗

bioRxiv2025

Highly potent novel multi-armoured IL13Rα2 CAR-T subverts the immunosuppressive microenvironment of Glioblastoma

Glioblastoma remains one of the most challenging and lethal brain cancers, with limited treatment options. While CAR-T cells have shown promise in some patients, sustaining T-cell activity and overcoming the immunosuppressive tumour microenvironment remain significant hurdles. Here, we present an armoured CAR-T cell design to address these challenges and enhance persistence in GBM tumours. We developed a highly specific humanised single-domain antibody targeting IL13R2 and included it alongside four additional modular elements in a single retroviral vector for CAR-T generation. Our results demonstrate that this single-cassette CAR-T cell design possesses high resilience against TGF-{beta}-mediated immunosuppression, enhanced tumour-killing capacity through IL-12 secretion while maintaining a favourable safety profile, extended persistence in the host, and an additional layer of safety control through the incorporation of a suicide switch. Importantly, despite its complexity, the construct can still be manufactured efficiently. These advancements represent a significant step forward in addressing key challenges associated with CAR-T cell therapy in solid tumours.

cancer biology↗

bioRxiv2024

Prediction of protein biophysical traits from limited data: a case study on nanobody thermostability through NanoMelt

1In-silico prediction of protein biophysical traits is often hindered by the limited availability of experimental data and their heterogeneity. Training on limited data can lead to overfitting and poor generalisability to sequences distant from those in the training set. Additionally, inadequate use of scarce and disparate data can introduce biases during evaluation, leading to unreliable model performances being reported. Here, we present a comprehensive study exploring various approaches for protein fitness prediction from limited data, leveraging pre-trained embeddings, repeated stratified nested cross-validation, and ensemble learning to ensure an unbiased assessment of the performances. We applied our framework to introduce NanoMelt, a predictor of nanobody thermostability trained with a dataset of 640 measurements of apparent melting temperature, obtained by integrating data from the literature with 129 new measurements from this study. We find that an ensemble model stacking multiple regression using diverse sequence embeddings achieves state-of-the-art accuracy in predicting nanobody thermostability. We further demonstrate NanoMelts potential to streamline nanobody development by guiding the selection of highly stable nanobodies. We make the curated dataset of nanobody thermostability freely available and NanoMelt accessible as a downloadable software and webserver. 2 Significance StatementRapidly predicting protein biophysical traits with accuracy is a key goal in protein engineering, yet efforts to develop reliable predictors are often hindered by limited and disparate experimental measurements. We introduce a framework to predict biophysical traits using few training data, leveraging diverse machine learning approaches via a semi-supervised framework combined with ensemble learning. We applied this framework to develop NanoMelt, a tool to predict nanobody thermostability trained on a new dataset of apparent melting temperatures. Nanobodies are increasingly important in research and therapeutics due to their ease of production and small size, which allows deeper tissue penetration and seamless combination into multi-specific compounds. NanoMelt outperforms available methods for protein thermostability prediction and can streamline nanobody development by guiding the design and selection of highly stable nanobodies during discovery and optimization campaigns.

bioengineering↗

bioRxiv2023

AbNatiV: VQ-VAE-based assessment of antibody and nanobody nativeness for engineering, selection, and computational design

Monoclonal antibodies have emerged as key therapeutics, and nanobodies are rapidly gaining momentum following the approval of the first nanobody drug in 2019. Nonetheless, the development of these biologics as therapeutics remains a challenge. Despite the availability of established in vitro directed evolution technologies that are relatively fast and cheap to deploy, the gold standard for generating therapeutic antibodies remains discovery from animal immunization or patients. Immune-system derived antibodies tend to have favourable properties in vivo, including long half-life, low reactivity with self-antigens, and low toxicity. Here, we present AbNatiV, a deep-learning tool for assessing the nativeness of antibodies and nanobodies, i.e., their likelihood of belonging to the distribution of immune-system derived human antibodies or camelid nanobodies. AbNatiV is a multi-purpose tool that accurately predicts the nativeness of Fv sequences from any source, including synthetic libraries and computational design. It provides an interpretable score that predicts the likelihood of immunogenicity, and a residue-level profile that can guide the engineering of antibodies and nanobodies indistinguishable from immune-system-derived ones. We further introduce an automated humanisation pipeline, which we applied to two nanobodies. Wet-lab experiments show that AbNatiV-humanized nanobodies retain binding and stability at par or better than their wild type, unlike nanobodies humanised relying on conventional structural and residue-frequency analysis. We make AbNatiV available as downloadable software and as a webserver.

bioengineering↗

bioRxiv2023

Sequence-based prediction of the solubility of peptides containing non-natural amino acids

Non-natural amino acids are increasingly used as building blocks in the development of peptide-based drugs, as they expand the available chemical space to tailor function, half-life and other key properties. However, while the chemical space of modified amino acids (mAAs) is potentially vast, experimental methods for measuring the developability properties of mAA-containing peptides are expensive and time consuming. To facilitate developability programs through computational methods, we present CamSol-PTM, a method that enables the fast and reliable sequence-based prediction of the solubility of mAA-containing peptides. From a computational screening of 50,000 mAA-containing variants of three peptides, we selected five different mAAs for a total number of 30 peptide variants for experimental validation. We demonstrate the accuracy of the predictions by comparing the calculated and experimental solubility values. Our results indicate that the computational screening of mAA-containing peptides can extend by over four orders of magnitude the ability to explore the solubility chemical space of peptides. This method is available as a web server at https://www-cohsoftware.ch.cam.ac.uk/index.php/camsolptm.

biophysics↗