Abstract:
Biological complexes have a fundamental role in cellular processes, and understanding their interactions at the atomic level is crucial for the advancment in structural biology and drug development. This thesis presents several significant contributions to the field of structural bioinformatics, focusing on characterizing interactions in biological complexes, particularly antibody/nanobody-antigen interfaces, through algotihms implementation and comprehensive analyses.
The first major contribution of the present work is the development of Iter-CONSRANK, an iterative consensus-based algorithm designed to improve the scoring of protein-protein docking models. We addresses one of the field's open challenges: identifying correct solutions within large ensembles of docking models. Iter-CONSRANK is an implementation of CONSRANK algorithm, consisting of an iterative process to discard lower-ranked models based on inter-residue contact conservation. Its erformance was extensively evaluated using two challenging datasets: the 3K-BM5up dataset, comprising around 156,000 models, and the CAPRI-derived Score_set, containing about 20,000 models. Results demonstrated significant improvement in the proportion of correct solutions among different model difficulty categories. When compared to other scoring functions, Iter-CONSRANK showed superior performance in ranking correct models at top positions, making it a valuable tool for both pre-processing docking ensembles and independent scoring.
The second major contribution is the creation and analysis of manually curated, non-redundant datasets of antibody-antigen and nanobody-antigen complexes. Two distinct datasets were established: one comprising 319 unique nanobody-antigen interfaces from 297 structures, and another containing 702 unique antibody-antigen interfaces from 679 structures. These datasets were derived using defined criteria, such as resolution cutoffs, sequence clustering with identity thresholds, different for antibodies/nanobodies sequence antigens one. The analysis revealed remarkable similarities in interfaces, despite significant size difference between antibodies and nanobodies. Both showed comparable interface area distributions and similar correlations between interface area and number of contacts, proving how nanobodies achieve binding affinities comparable to traditional antibodies.
Another significant contribution is the development of computational tools for analyzing atomic-level interactions at protein interfaces. The in-house pipeline and script shaped for this work allows comprehensive characterization of various atomic interaction types, including Hydrogen bonds, salt-bridges, π-π stacking interactions, cation-π interactions, anion-π interactions, polar-π interactions, water-mediated contacts, metal-mediated contacts, and van der Waals interactions.
The analysis of the datasets lead several important discoveries regarding sequence-structure relationships in immunoglobulins. Of particular interest was the discovery of the previously unrecognized importance of serine in nanobody CDR2, showing high conservation at specific positions and playing a crucial role in forming polar-vdW interactions and water-mediated contacts. The analysis also highlighted the significant contribution of framework regions to binding interfaces, challenging the traditional focus on CDRs alone.
Moreover, analysis revealed that despite nanobodies are significantly smaller than antibodies, they achieve similar binding abilities, due to comparable interface areas/interactions ratio. The most frequent atomic interaction type in both cases was found to be polar-vdW, followed by apolar-vdW and CH-O/N bonds. Notable differences between nanobodies and antibodies were observed in the distribution of specific interaction types, such as cation-π and anion-π interactions being slightly more abundant in nanobodies, while π-π stacking and polar-π interactions were more prevalent in antibodies.
The advances, comprehensive datasets, and novel insights provided by this work establish a foundation for future research in structural biology, protein engineering, and drug development. As this field continues to advance and evolve, particularly with developments artificial intelligence-based structure prediction, these contributions will become increasingly valuable for understanding and engineering protein-protein interactions. The comprehensive nature of the analysis provides a template for future studies of other protein-protein interfaces, while the specific findings about immunoglobulin interfaces could be at the base for development of next-generation therapeutics.

Bio:
Tiziana has obtained her bachelor's degree in Biology at University of Napoli "Parthenope" and her master's degree in Bioinformatics at University of Bologna. Under the supervision of Prof. Luigi Cavallo and Prof. Romina Oliva, she focused her research on atomic interactions at protein complexes interface. Specifically, she characterized antibody/nanobody-antigen interfaces, manually curating a large non-redundant dataset. Moreover she was chair of the European Student Council Symposium (ESCS 2024) in Turku - Norway, as she is an active member of the International Society of Computational Biology (ISCB). Lately, he has been nominated as chair of symposia for the ISCB Student Council, and she is currently part of 3D-
SIG community of ELIXIR Europe and CASP conformational ensemble Special interest Group.