Artificial-Intelligence-based Methods for Nanobody Design – ANDES

ANDES aims at developing and experimentally validating new computational methods for nanobody design, with a large spectrum of applications (therapies, diagnosis…) in order to have efficient technological solutions, ready to react very quickly and to face the emergence of new viruses. Our ambition lies in the explicit modeling of crucial features for the design of functional nanobodies: binding specificity and affinity, dense buried polar interaction networks, binding-competent loop conformations. A first development will consist in designing and training a new family of decomposable energy (score) functions targeted at the design of protein-protein interfaces. These functions will be deep learned from structural interfaces and should avoid the usual weakness of existing design functions that tend to design interfaces as tight hydrophobic cores while both sides of the interface have to exist also in the solvent, when unbound. Developments will then aim to interleave the explorations of the conformation space and of the sequence space, while integrating machine-learned information extracted from known sequences and structures. We will exploit efficient robotics-inspired methods and automated reasoning from AI, combined with molecular modeling approaches. In terms of sequence design, the automated reasoning technology we use for design will be extended to favor the design of sequences that promote sequences stabilizing a single loop conformation (and not others), the conformation which is coimpetent for binding. The expoected efficiency of the energy function and the established efficiency of both the automated-reasoning based design technology as well as robotics-inspired conformation design technology we use will be essential to simultaneously explore sequence and conformation spaces. Nanobodies targeting coronavirus antigens will be thus designed for specificity changing and affinity optimization. In silico designed nanobodies will be synthesized and experimentally validated by biophysical and structural biology methods all along the project, allowing feedback on the precise criteria that in silico design should target. By establishing the proof-of-concept of an efficient and reliable computational nanobody design framework, the interdisciplinary project, ANDES, will : (i) accelerate the development of innovative nanobody-based therapies and diagnostics; (ii) bring highly valuable predictive tools for the tailor-made conception of biomolecules of interest for numerous biotechnology and health fields; (iii) provide advanced AI and computational biology algorithms and methods of general interest; (iv) contribute to improve our understanding of protein interactions, the role of loops at the binding interface and protein sequence-structure-function relationships.