Pietro Sormanni

This paper predicts antibody non-specificity using protein language models and biophysical parameters, showing that heavy variable domains and isoelectric point are key predictors.

This paper introduces AbNatiV2 and p-AbNatiV2, deep learning models for assessing nativeness and pairing likelihood in antibodies and nanobodies, showing improved performance on held-out tests.

This paper develops an in vivo CAR T cell platform using lipid nanoparticles and T-trex strategy to enhance safety, efficacy, and tumor control.

The paper evaluates AI-based antibody-antigen structure prediction methods, finding that confidence scores often fail to distinguish correct from incorrect pairings, even with increased sampling.

This paper introduces NbForge, a nanobody folding model that improves prediction of HCDR3 blueprint and disulphide bonds, achieving accuracy comparable to large predictors with faster speed.

The paper quantifies how different excipients affect the solubility of therapeutic antibodies using a high-throughput platform and identifies molecular-specific and nonspecific interactions.

SurfDiff predicts epitope selectivity and cross-reactivity by analyzing protein surface features, outperforming existing metrics in binder and immunogen design.

This study improves nanobody structure prediction by using synthetic data and a classifier to filter structural accuracy, achieving better results than existing methods with lower computational cost.

This paper shows that BLOSUM similarity scores correlate with antibody binding affinity, suggesting generative models implicitly learn evolutionary priors.

The paper analyzes 3,766 binders to identify predictive features for experimental success in de novo binder design, showing that AF3-derived metrics improve prediction accuracy.

This study develops and characterizes humanized TNFα inhibitory nanobodies with picomolar potency and low immunogenicity for therapeutic use in TNFα-mediated diseases.

This paper develops humanized TfR1-targeting nanobodies (NewroBus) that cross the BBB and deliver therapeutic payloads to the CNS.

This study identifies two critical sites in the C-terminal β-chain tail of clusterin that mediate its interaction with misfolded proteins, including amyloid β, through experimental and computational approaches.

This paper develops a novel CAR-T cell therapy for glioblastoma that overcomes immunosuppression and improves persistence through a multi-armoured design.

NanoMelt, a ridge-stacked ensemble of regressors on four diverse sequence embeddings, predicts nanobody melting temperature from 640 curated sequences, reaching Spearman 0.832 under nested cross-validation.

The paper introduces SpyBLI, a cell-free pipeline for rapid, accurate binding kinetics measurement from crude samples, demonstrating its applicability with nanobodies and scFvs.

A fragment-based computational method designs nanobodies targeting specific epitopes with nanomolar affinities, without requiring high-resolution antigen structures.

The paper evaluates 43 methods for predicting intrinsically disordered regions and binding regions in proteins, finding that deep learning methods outperform older physicochemical approaches but binding prediction remains challenging.

The paper identifies amino acid pairs at the ribosome's P- and A-sites that predictably and causally modulate translation-elongation rates in yeast.

The paper develops chemical kinetic methods to measure translation initiation, elongation, and codon-specific rates from ribosome profiling data, finding significant variability in codon translation rates.

This paper uses Integer Programming to accurately identify A-site locations on ribosome-protected mRNA fragments from Ribo-Seq data, improving signal-to-noise ratio by utilizing a wider range of fragment sizes.

This paper introduces Parapred, a deep learning model for paratope prediction that outperforms existing methods using only amino acid sequence data.