Data Fusion

Several experimental techniques are aimed at studying protein structure, which in turn can provide precious insights about molecular functions. Data are often rationalized against available high resolution structures obtained by X-ray crystallography or NMR. This can be problematic because (1) the protein conformation in an experiment might differ from its reference atomistic structure, (2) the data might only be explained by multiple conformations, and (3) the data obtained through different experiments might be ambiguous. In order to tackle these issues, we rely on sampling of proteins’ conformational space by means of molecular simulations and deep learning methods.


Particular attention is dedicated to the analysis and comparison of ion mobility, SAXS, and chemical cross-linking. Our aims are to:

  • develop a framework to assess experimental data against a full protein conformational ensemble and determine a conformational subset consistent with it
  • analyse whether, and to which degree, combinations of different techniques can enhance structure discrimination within an ensemble of candidates