data-evo-intermediate-reconstruction

This dataset contains the data associated with the work by Roberto Netti and Martin Weigt on Reconstructability of evolutionary intermediates in epistatic sequence landscapes. The study develops a controlled in silico framework to assess how much information about an evolutionary intermediate sequence can be recovered from its two endpoint sequences. Using family-specific generative landscapes inferred via Boltzmann Machine Direct Coupling Analysis (bmDCA), nucleotide-level evolutionary simulations, and autoregressive conditional modeling (arDCA), the work benchmarks different reconstruction strategies and characterizes their fundamental limits. The dataset includes multiple sequence alignments (MSAs), inferred bmDCA model parameters, simulated evolutionary trajectory triplets (S_start, S_mid, S_end), and trained arDCA model weights for the three protein families analyzed: Chorismate Mutase, β-lactamase (PF13354), and the Response Regulator domain (PF00072). It further includes precomputed Context-Dependent Entropy (CDE) profiles used for sequence mutability estimation and evolutionary timescale inference. The associated code for generative modeling, evolutionary simulations, and intermediate reconstruction is available at: https://github.com/robertonetti/evo-intermediate-reconstruction.git and https://github.com/robertonetti/arDCA-evo-intermediate-reconstruction.git