Deep Learning and Likelihood Approaches for Viral Phylogeography Converge on the Same Answers Whether the Inference Model Is Right or Wrong.

Analysis of phylogenetic trees has become an essential tool in epidemiology. Likelihood-based methods fit models to phylogenies to draw inferences about the phylodynamics and history of viral transmission. However, these methods are often computationally expensive, which limits the complexity and realism of phylodynamic models and makes them ill-suited for informing policy decisions in real-time during rapidly developing outbreaks. Likelihood-free methods using deep learning are pushing the boundaries of inference beyond these constraints. In this paper, we extend, compare, and contrast a recently developed deep learning method for likelihood-free inference from trees. We trained multiple deep neural networks using phylogenies from simulated outbreaks that spread among 5 locations and found they achieve close to the same levels of accuracy as Bayesian inference under the true simulation model. We compared robustness to model misspecification of a trained neural network to that of a Bayesian method. We found that both models had comparable performance, converging on similar biases. We also implemented a method of uncertainty quantification called conformalized quantile regression that we demonstrate has similar patterns of sensitivity to model misspecification as Bayesian highest posterior density (HPD) and greatly overlap with HPDs, but have lower precision (more conservative). Finally, we trained and tested a neural network against phylogeographic data from a recent study of the SARS-Cov-2 pandemic in Europe and obtained similar estimates of region-specific epidemiological parameters and the location of the common ancestor in Europe. Along with being as accurate and robust as likelihood-based methods, our trained neural networks are on average over 3 orders of magnitude faster after training. Our results support the notion that neural networks can be trained with simulated data to accurately mimic the good and bad statistical properties of the likelihood functions of generative phylogenetic models.

Ancestral reconstruction of protein interaction networks.

The molecular and cellular basis of novelty is an active area of research in evolutionary biology. Until very recently, the vast majority of cellular phenomena were so difficult to sample that cross-species studies of biochemistry were rare and comparative analysis at the level of biochemical systems was almost impossible. Recent advances in systems biology are changing what is possible, however, and comparative phylogenetic methods that can handle this new data are wanted. Here, we introduce the term "phylogenetic latent variable models" (PLVMs, pronounced "plums") for a class of models that has recently been used to infer the evolution of cellular states from systems-level molecular data, and develop a new parameterization and fitting strategy that is useful for comparative inference of biochemical networks. We deploy this new framework to infer the ancestral states and evolutionary dynamics of protein-interaction networks by analyzing >16,000 predominantly metazoan co-fractionation and affinity-purification mass spectrometry experiments. Based on these data, we estimate ancestral interactions across unikonts, broadly recovering protein complexes involved in translation, transcription, proteostasis, transport, and membrane trafficking. Using these results, we predict an ancient core of the Commander complex made up of CCDC22, CCDC93, C16orf62, and DSCR3, with more recent additions of COMMD-containing proteins in tetrapods. We also use simulations to develop model fitting strategies and discuss future model developments.

Towards Consensus Gene Ages.

Correctly estimating the age of a gene or gene family is important for a variety of fields, including molecular evolution, comparative genomics, and phylogenetics, and increasingly for systems biology and disease genetics. However, most studies use only a point estimate of a gene's age, neglecting the substantial uncertainty involved in this estimation. Here, we characterize this uncertainty by investigating the effect of algorithm choice on gene-age inference and calculate consensus gene ages with attendant error distributions for a variety of model eukaryotes. We use 13 orthology inference algorithms to create gene-age datasets and then characterize the error around each age-call on a per-gene and per-algorithm basis. Systematic error was found to be a large factor in estimating gene age, suggesting that simple consensus algorithms are not enough to give a reliable point estimate. We also found that different sources of error can affect downstream analyses, such as gene ontology enrichment. Our consensus gene-age datasets, with associated error terms, are made fully available at so that researchers can propagate this uncertainty through their analyses (geneages.org).

Conserved properties of individual Ca2+-binding sites in calmodulin.

Calmodulin (CaM) is a Ca(2+)-sensing protein that is highly conserved and ubiquitous in eukaryotes. In humans it is a locus of life-threatening cardiomyopathies. The primary function of CaM is to transduce Ca(2+) concentration into cellular signals by binding to a wide range of target proteins in a Ca(2+)-dependent manner. We do not fully understand how CaM performs its role as a high-fidelity signal transducer for more than 300 target proteins, but diversity among its four Ca(2+)-binding sites, called EF-hands, may contribute to CaM's functional versatility. We therefore looked at the conservation of CaM sequences over deep evolutionary time, focusing primarily on the four EF-hand motifs. Expanding on previous work, we found that CaM evolves slowly but that its evolutionary rate is substantially faster in fungi. We also found that the four EF-hands have distinguishing biophysical and structural properties that span eukaryotes. These results suggest that all eukaryotes require CaM to decode Ca(2+) signals using four specialized EF-hands, each with specific, conserved traits. In addition, we provide an extensive map of sites associated with target proteins and with human disease and correlate these with evolutionary sequence diversity. Our comprehensive evolutionary analysis provides a basis for understanding the sequence space associated with CaM function and should help guide future work on the relationship between structure, function, and disease.

Complex Homology and the Evolution of Nervous Systems.

We examine the complex evolution of animal nervous systems and discuss the ramifications of this complexity for inferring the nature of early animals. Although reconstructing the origins of nervous systems remains a central challenge in biology, and the phenotypic complexity of early animals remains controversial, a compelling picture is emerging. We now know that the nervous system and other key animal innovations contain a large degree of homoplasy, at least on the molecular level. Conflicting hypotheses about early nervous system evolution are due primarily to differences in the interpretation of this homoplasy. We highlight the need for explicit discussion of assumptions and discuss the limitations of current approaches for inferring ancient phenotypic states.

Applications of comparative evolution to human disease genetics.

Direct comparison of human diseases with model phenotypes allows exploration of key areas of human biology which are often inaccessible for practical or ethical reasons. We review recent developments in comparative evolutionary approaches for finding models for genetic disease, including high-throughput generation of gene/phenotype relationship data, the linking of orthologous genes and phenotypes across species, and statistical methods for linking human diseases to model phenotypes.

Convergence of ion channel genome content in early animal evolution.

Multicellularity has evolved multiple times, but animals are the only multicellular lineage with nervous systems. This fact implies that the origin of nervous systems was an unlikely event, yet recent comparisons among extant taxa suggest that animal nervous systems may have evolved multiple times independently. Here, we use ancestral gene content reconstruction to track the timing of gene family expansions for the major families of ion-channel proteins that drive nervous system function. We find that animals with nervous systems have broadly similar complements of ion-channel types but that these complements likely evolved independently. We also find that ion-channel gene family evolution has included large loss events, two of which were immediately followed by rounds of duplication. Ctenophores, cnidarians, and bilaterians underwent independent bouts of gene expansion in channel families involved in synaptic transmission and action potential shaping. We suggest that expansions of these family types may represent a genomic signature of expanding nervous system complexity. Ancestral nodes in which nervous systems are currently hypothesized to have originated did not experience large expansions, making it difficult to distinguish among competing hypotheses of nervous system origins and suggesting that the origin of nerves was not attended by an immediate burst of complexity. Rather, the evolution of nervous system complexity appears to resemble a slow fuse in stem animals followed by many independent bouts of gene gain and loss.

Evolution of voltage-gated ion channels at the emergence of Metazoa.

Voltage-gated ion channels are large transmembrane proteins that enable the passage of ions through their pore across the cell membrane. These channels belong to one superfamily and carry pivotal roles such as the propagation of neuronal and muscular action potentials and the promotion of neurotransmitter secretion in synapses. In this review, we describe in detail the current state of knowledge regarding the evolution of these channels with a special emphasis on the metazoan lineage. We highlight the contribution of the genomic revolution to the understanding of ion channel evolution and for revealing that these channels appeared long before the appearance of the first animal. We also explain how the elucidation of channel selectivity properties and function in non-bilaterian animals such as cnidarians (sea anemones, corals, jellyfish and hydroids) can contribute to the study of channel evolution. Finally, we point to open questions and future directions in this field of research.

Ancient association between cation leak channels and Mid1 proteins is conserved in fungi and animals.

Neuronal resting potential can tune the excitability of neural networks, affecting downstream behavior. Sodium leak channels (NALCN) play a key role in rhythmic behaviors by helping set, or subtly changing neuronal resting potential. The full complexity of these newly described channels is just beginning to be appreciated, however. NALCN channels can associate with numerous subunits in different tissues and can be activated by several different peptides and second messengers. We recently showed that NALCN channels are closely related to fungal calcium channels, which they functionally resemble. Here, we use this relationship to predict a family of NALCN-associated proteins in animals on the basis of homology with the yeast protein Mid1, the subunit of the yeast calcium channel. These proteins all share a cysteine-rich region that is necessary for Mid1 function in yeast. We validate this predicted association by showing that the Mid1 homolog in Drosophila, encoded by the CG33988 gene, is coordinately expressed with NALCN, and that knockdown of either protein creates identical phenotypes in several behaviors associated with NALCN function. The relationship between Mid1 and leak channels has therefore persisted over a billion years of evolution, despite drastic changes to both proteins and the organisms in which they exist.

Phylogeny unites animal sodium leak channels with fungal calcium channels in an ancient, voltage-insensitive clade.

Proteins in the superfamily of voltage-gated ion channels mediate behavior across the tree of life. These proteins regulate the movement of ions across cell membranes by opening and closing a central pore that controls ion flow. The best-known members of this superfamily are the voltage-gated potassium, calcium (Ca(v)), and sodium (Na(v)) channels, which underlie impulse conduction in nerve and muscle. Not all members of this family are opened by changes in voltage, however. NALCN (NA(+) leak channel nonselective) channels, which encode a voltage-insensitive "sodium leak" channel, have garnered a growing interest. This study examines the phylogenetic relationship among Na(v)/Ca(v) voltage-gated and voltage-insensitive channels in the eukaryotic group Opisthokonta, which includes animals, fungi, and their unicellular relatives. We show that NALCN channels diverged from voltage-gated channels before the divergence of fungi and animals and that the closest relatives of NALCN channels are fungal calcium channels, which they functionally resemble.

Evolution of sodium channels predates the origin of nervous systems in animals.

Voltage-dependent sodium channels are believed to have evolved from calcium channels at the origin of the nervous system. A search of the genome of a single-celled choanoflagellate (the sister group of animals) identified a gene that is homologous to animal sodium channels and has a putative ion selectivity filter intermediate between calcium and sodium channels. Searches of a wide variety of animal genomes, including representatives of each basal lineage, revealed that similar homologs were retained in most lineages. One of these, the Placozoa, does not possess a nervous system. We cloned and sequenced the full choanoflagellate channel and parts of two placozoan channels from mRNA, showing that they are expressed. Phylogenetic analysis clusters the genes for these channels with other known sodium channels. From this phylogeny we infer ancestral states of the ion selectivity filter and show that this state has been retained in the choanoflagellate and placozoan channels. We also identify key gene duplications and losses and show convergent amino acid replacements at important points along the animal lineage.

Evolution of sodium channels and the new view of early nervous system evolution.

Advances in genomics have revealed that many genes implicated in the nervous systems of bilaterians were already present in the last common ancestor (LCA) of animals, and some even before that.(1) (-) (5) This new information coincides with a growing reinterpretation of cnidarian nervous systems which holds that they are 'fundamentally conventional' with regards to bilaterian nervous systems,(6) and do not represent ancient forms. Since in general adult forms are expected to be the most derived features of organisms, the study of non-bilaterian larval forms may be a better way to investigate potential plesiomorphies. We recently showed that voltage-gated sodium channel (Na(v)) genes, which make action potentials in nerves and muscles, were present in the LCA of animals and choanoflagellates, the closest unicellular relatives to animals.(2) This addendum will attempt to put this finding within the context of the new views of nervous system evolution.