Comparative analysis of structure and sequence based methods for the prediction of linear motif binding energies.

GLAVINA JULIANA; ALVAREZ, LUCIA; FERNANDEZ-BALLESTER GREGORIO; SEHR PETER; LEWIS JOE; GIBSON TOBY J; SANCHEZ IGNACIO E; CHEMES LUCIA BEATRIZ

Mendoza

Conferencia; 10mo Congreso Argentino de Bioinformática y Biologı́a Computacional.; 2019

Asociacion Argentina de Bioinformática y Biología Computacional

BACKGROUND: Linear motifs (LMs) are ubiquitous short sequence elements that participate i nprotein-protein i nteractions. The LxCxE motif i s a 9-residue sequence that i nteracts with theretinoblastoma protein (pRb), which plays a key role i n cell cycle progression. Viruses mimic theLxCxE motif to hijack pRb i nducing proliferation and thus promoting the synthesis of the viralgenome. However, the mechanism by which viruses outcompete the host i nteractions i s stillunclear. One main hypothesis i s that viral motifs evolve higher affinities than cellular motifs. Apredictor that represents LMs mediated i nteractions properly would facilitate the understanding ofthese mechanisms. Here, we used sequence- and structure-based approaches to obtainpredictions of binding affinities of cellular, viral and mutant sequences of the LxCxE motif, whichallowed us to compare different methods and describe their reliability i n predicting the true bindingaffinities measured from the sequences.RESULTS: Structure-based calculations used the empirical force field FoldX under differentconditions. Using available LxCxE-pRb structures, we used FoldX to study the binding energy ofeach LxCxE peptide to pRb. Sequence-based calculations used molecular i nformation theory,which makes use of residue statistics at an alignment of known LMs. We used an alignment ofLxCxE motifs to compute a position specific scoring matrix. The different methods were able toreproduce the experimental data with different accuracy. For i nstance, we found major differencesat a conserved hydrophobic position (L28 i n the E7 protein) after the LxCxE core that associates toa pRb groove. Experimental data i ndicated the broad hydrophobic groove i n pRb accommodatesa wide range of amino acids. However, the marginal decrease i n affinity (0.6x) of the L28Wmutation was poorly captured by all structure-based calculations, where the tryptophan was highlypenalized in that position.DISCUSSION: The structure-based methods reflected almost all the sequence variability i n thedataset but they are biased. The sequence-based method reflected the sequence variability, butthey were highly dependent on the i nitial dataset and did not reflect the structural flexibility.However, both approaches were able to reproduce quantitative binding experiments to anacceptable degree, suggesting they constitute useful tools to predict LMs binding energetics.