The jigsaw puzzle of sequence phenotype inference: Piecing together Shannon entropy, importance sampling, and Empirical Bayes

A model of k-mer surprisal to quantify local sequence information content surrounding splice regions

Molecular sequences carry information. Analysis of sequence conservation between homologous loci is a proven approach with which to explore the information content of molecular sequences. This is often done using multiple sequence alignments to support comparisons between homologous loci. These methods therefore rely on sufficient underlying sequence similarity with which to construct a representative alignment. Here we describe a method using a formal metric of information, surprisal, to analyse biological sub-sequences without alignment constraints. We applied our model to the genomes of five different species to reveal similar patterns across a panel of eukaryotes. As the surprisal of a sub-sequence is inversely proportional to its occurrence within the genome, the optimal size of the sub-sequences was selected for each species under consideration. With the model optimized, we found a strong correlation between surprisal and CG dinucleotide usage. The utility of our model was tested by examining the sequences of genes known to undergo splicing. We demonstrate that our model can identify biological features of interest such as known donor and acceptor sites. Analysis across all annotated coding exon junctions in Homo sapiens reveals the information content of coding exons to be greater than the surrounding intron regions, a consequence of increased suppression of the CG dinucleotide in intronic space. Sequences within coding regions proximal to exon junctions exhibited novel patterns within DNA and coding mRNA that are not a function of the encoded amino acid sequence. Our findings are consistent with the presence of secondary information encoding features such as DNA and RNA binding sites, multiplexed through the coding sequence and independent of the information required to define the corresponding amino-acid sequence. We conclude that surprisal provides a complementary methodology with which to locate regions of interest in the genome, particularly in situations that lack an appropriate multiple sequence alignment.

Block network mapping approach to quantitative trait locus analysis

Background

Advances in experimental biology have enabled the collection of enormous troves of data on genomic variation in living organisms. The interpretation of this data to extract actionable information is one of the keys to developing novel therapeutic strategies to treat complex diseases. Network organization of biological data overcomes measurement noise in several biological contexts. Does a network approach, combining information about the linear organization of genomic markers with correlative information on these markers in a Bayesian formulation, lead to an analytic method with higher power for detecting quantitative trait loci?

Results

Block Network Mapping, combining Similarity Network Fusion (Wang et al., NM 11:333–337, 2014) with a Bayesian locus likelihood evaluation, leads to large improvements in area under the receiver operating characteristic and power over interval mapping with expectation maximization. The method has a monotonically decreasing false discovery rate as a function of effect size, unlike interval mapping.

Conclusions

Block Network Mapping provides an alternative data-driven approach to mapping quantitative trait loci that leverages correlations in the sampled genotypes. The evaluation methodology can be combined with existing approaches such as Interval Mapping. Python scripts are available at http://lbm.niddk.nih.gov/vipulp/. Genotype data is available at http://churchill-lab.jax.org/website/GattiDOQTL.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-016-1351-8) contains supplementary material, which is available to authorized users.

Correlated rigid modes in protein families

A great deal of evolutionarily conserved information is contained in genomes and proteins. Enormous effort has been put into understanding protein structure and developing computational tools for protein folding, and many sophisticated approaches take structure and sequence homology into account. Several groups have applied statistical physics approaches to extracting information about proteins from sequences alone. Here, we develop a new method for sequence analysis based on first principles, in information theory, in statistical physics and in Bayesian analysis. We provide a complete derivation of our approach and we apply it to a variety of systems, to demonstrate its utility and its limitations. We show in some examples that phylogenetic alignments of amino-acid sequences of families of proteins imply the existence of a small number of modes that appear to be associated with correlated global variation. These modes are uncovered efficiently in our approach by computing a non-perturbative effective potential directly from the alignment. We show that this effective potential approaches a limiting form inversely with the logarithm of the number of sequences. Mapping symbol entropy flows along modes to underlying physical structures shows that these modes arise due to correlated compensatory adjustments. In the protein examples, these occur around functional binding pockets.