Information-theoretic analysis and prediction of protein atomic burials: on the search for an informational intermediate between sequence and structure

Abstract

The dataset of protein structures and the prediction implementations are available at http://www.btc.unb.br/ (in 'Software').

Information and redundancy in the burial folding code of globular proteins within a wide range of shapes and sizes

“…It appears natural, in this context, to consider specific protein sequences just as signals being transmitted through a communication channel. The fact that they are hard to distinguish from random sequences suggests an efficient encoding of structural messages with the appropriate average information content, or entropy, resulting in a rate of input information close to the maximal for 20 uncorrelated symbols distributed according to observed amino acid frequencies, that is ≈4.1 bits/residue or ≈0.5 bits/atom. It is a reasonable possibility that an appropriate folding potential could be parameterized in terms of these hypothetical nonrandom structural messages that, even though encoded in sequences, would be more directly correlated with tertiary structures.…”

“…We have been exploring the possibility that atomic burials, as quantified by central distances, could constitute such messages . In terms of a simple analogy with human communication, which is actually similar to the one discussed in Ref.…”

“…The average information content of burial sequences, estimated from burial local correlations in small globular proteins for small L , was additionally found to be comparable to the entropy of protein sequences and therefore sufficiently small to be at least in principle further encoded in them. Burial information was later shown to be actually extractable from sequence by simple statistical prediction schemes, providing an estimated average reduction in burial uncertainty around 15% for 2 or 3 burial layers of C α or C β atoms, suggesting that the resulting sequence‐dependent burial predictions could be appropriate for folding simulations if the burial folding code turned out to be sufficiently redundant. Recent ab initio simulations using discrete burial information obtained directly from sequence with a burial prediction algorithm based on a Hidden Markov Model (HMM), which was adapted from an algorithm for secondary structure prediction, have been encouraging in this respect.…”

“…Burial information was later shown to be actually extractable from sequence by simple statistical prediction schemes, providing an estimated average reduction in burial uncertainty around 15% for 2 or 3 burial layers of C α or C β atoms, suggesting that the resulting sequence‐dependent burial predictions could be appropriate for folding simulations if the burial folding code turned out to be sufficiently redundant. Recent ab initio simulations using discrete burial information obtained directly from sequence with a burial prediction algorithm based on a Hidden Markov Model (HMM), which was adapted from an algorithm for secondary structure prediction, have been encouraging in this respect. Native‐like conformations were obtained and identified for three small globular proteins, comprising all three structural classes, with an above average burial prediction accuracy around 56% for L = 4 burial layers .…”

Quantitative theory of hydrophobic effect as a driving force of protein structure

“…The model can also provide auxilliary information for MD simulations that use burial traces to generate initial protein configurations. 36 In addition, because of the high speed, the model can be used as a tool to study large collections of homologous sequences, which became available with high-throughput genomic sequencing and to access structural information about different mutants that are not yet crystallized.…”