Inferring divergence of context-dependent substitution rates in Drosophila genomes with applications to comparative genomics

Learning Context-Dependent DNA Mutation Patterns in Error-Prone Polymerase Chain Reaction

We present a novel statistical learning method for studying context-dependent error rates in error-prone polymerase chain reaction (PCR) experiments. We demonstrate the method by applying it to error-prone PCR sequencing data and show how it may be exploited to improve the evolvability of genes in protein engineering.

EvoLSTM: context-dependent models of sequence evolution using a sequence-to-sequence LSTM

Motivation

Accurate probabilistic models of sequence evolution are essential for a wide variety of bioinformatics tasks, including sequence alignment and phylogenetic inference. The ability to realistically simulate sequence evolution is also at the core of many benchmarking strategies. Yet, mutational processes have complex context dependencies that remain poorly modeled and understood.

Results

We introduce EvoLSTM, a recurrent neural network-based evolution simulator that captures mutational context dependencies. EvoLSTM uses a sequence-to-sequence long short-term memory model trained to predict mutation probabilities at each position of a given sequence, taking into consideration the 14 flanking nucleotides. EvoLSTM can realistically simulate mammalian and plant DNA sequence evolution and reveals unexpectedly strong long-range context dependencies in mutation probabilities. EvoLSTM brings modern machine-learning approaches to bear on sequence evolution. It will serve as a useful tool to study and simulate complex mutational processes.

Availability and implementation

Code and dataset are available at https://github.com/DongjoonLim/EvoLSTM.

Supplementary information

Supplementary data are available at Bioinformatics online.

Net Evolutionary Loss of Residue Polarity in Drosophilid Protein Cores Indicates Ongoing Optimization of Amino Acid Composition

Amino acid frequencies in proteins may not be at equilibrium. We consider two possible explanations for the nonzero net residue fluxes in drosophilid proteins. First, protein interiors may have a suboptimal residue composition and be under a selective pressure favoring stability, that is, leading to the loss of polar (and the gain of large) amino acids. One would then expect stronger net fluxes on the protein interior than at the exposed sites. Alternatively, if most of the polarity loss occurs at the exposed sites and the selective constraint on amino acid composition at such sites decreases over time, net loss of polarity may be neutral and caused by disproportionally high occurrence of polar residues at exposed, least constrained sites. We estimated net evolutionary fluxes of residue polarity and volume at sites with different solvent accessibility in conserved protein families from 12 species of Drosophila. Net loss of polarity, miniscule in magnitude, but consistent across all lineages, occurred at all sites except the most exposed ones, where net flux of polarity was close to zero or, in membrane proteins, even positive. At the intermediate solvent accessibility the net fluxes of polarity and volume were similar to neutral predictions, whereas much of the polarity loss not attributable to neutral expectations occurred at the buried sites. These observations are consistent with the hypothesis that residue composition in many proteins is structurally suboptimal and continues to evolve toward lower polarity in the protein interior, in particular in proteins with intracellular localization. The magnitude of polarity and volume changes was independent from the protein’s evolutionary age, indicating that the approach to equilibrium has been slow or that no such single equilibrium exists.

Cooperativity, specificity, and evolutionary stability of Polycomb targeting in Drosophila

Metazoan genomes are partitioned into modular chromosomal domains containing active or repressive chromatin. In flies, Polycomb group (PcG) response elements (PREs) recruit PHO and other DNA-binding factors and act as nucleation sites for the formation of Polycomb repressive domains. The sequence specificity of PREs is not well understood. Here, we use comparative epigenomics and transgenic assays to show that Drosophila domain organization and PRE specification are evolutionarily conserved despite significant cis-element divergence within Polycomb domains, whereas cis-element evolution is strongly correlated with transcription factor binding divergence outside of Polycomb domains. Cooperative interactions of PcG complexes and their recruiting factor PHO stabilize PHO recruitment to low-specificity sequences. Consistently, PHO recruitment to sites within Polycomb domains is stabilized by PRC1. These data suggest that cooperative rather than hierarchical interactions among low-affinity sequences, DNA-binding factors, and the Polycomb machinery are giving rise to specific and strongly conserved 3D structures in Drosophila.

Drosophila functional elements are embedded in structurally constrained sequences

Modern functional genomics uncovered numerous functional elements in metazoan genomes. Nevertheless, only a small fraction of the typical non-exonic genome contains elements that code for function directly. On the other hand, a much larger fraction of the genome is associated with significant evolutionary constraints, suggesting that much of the non-exonic genome is weakly functional. Here we show that in flies, local (30–70 bp) conserved sequence elements that are associated with multiple regulatory functions serve as focal points to a pattern of punctuated regional increase in G/C nucleotide frequencies. We show that this pattern, which covers a region tenfold larger than the conserved elements themselves, is an evolutionary consequence of a shift in the balance between gain and loss of G/C nucleotides and that it is correlated with nucleosome occupancy across multiple classes of epigenetic state. Evidence for compensatory evolution and analysis of SNP allele frequencies show that the evolutionary regime underlying this balance shift is likely to be non-neutral. These data suggest that current gaps in our understanding of genome function and evolutionary dynamics are explicable by a model of sparse sequence elements directly encoding for function, embedded into structural sequences that help to define the local and global epigenomic context of such functional elements.

A key challenge in functional genomics is to predict evolutionary dynamics from functional annotation of the genome and vice versa. Modern epigenomic studies helped assign function to numerous new sequence elements, but left most of the genome essentially uncharacterized. Evolutionary genomics, on the other hand, consistently suggests that a much larger fraction of the un-annotated genome evolves under selective pressure. We hypothesize that this function-selection gap can be attributed to sequences that facilitate the physical organization of functional elements, such as transcription factor binding sites, within chromosomes. We exemplify this by studying in detail the sequences embedding small conserved elements (CEs) in Drosophila. We show that, while CEs have typically high AT content, high GC content levels around them are maintained by a non-neutral evolutionary balance between gain and loss of GC nucleotides. This non-uniform pattern is highly correlated with nucleosome organization around CEs, potentially imposing an evolutionary constraint on as much as one quarter of the genome. We suggest this can at least partly explain the above function-selection gap. Weak evolutionary constraints on “structural” sequences (at scales ranging from one nucleosome to recently described multi-megabase topological domains) may affect genome evolution just like structural motifs shape protein evolution.