Neural networks enable efficient and accurate simulation-based inference of evolutionary parameters from adaptation dynamics

The rate of adaptive evolution depends on the rate at which beneficial mutations are introduced into a population and the fitness effects of those mutations. The rate of beneficial mutations and their expected fitness effects is often difficult to empirically quantify. As these 2 parameters determine the pace of evolutionary change in a population, the dynamics of adaptive evolution may enable inference of their values. Copy number variants (CNVs) are a pervasive source of heritable variation that can facilitate rapid adaptive evolution. Previously, we developed a locus-specific fluorescent CNV reporter to quantify CNV dynamics in evolving populations maintained in nutrient-limiting conditions using chemostats. Here, we use CNV adaptation dynamics to estimate the rate at which beneficial CNVs are introduced through de novo mutation and their fitness effects using simulation-based likelihood–free inference approaches. We tested the suitability of 2 evolutionary models: a standard Wright–Fisher model and a chemostat model. We evaluated 2 likelihood-free inference algorithms: the well-established Approximate Bayesian Computation with Sequential Monte Carlo (ABC-SMC) algorithm, and the recently developed Neural Posterior Estimation (NPE) algorithm, which applies an artificial neural network to directly estimate the posterior distribution. By systematically evaluating the suitability of different inference methods and models, we show that NPE has several advantages over ABC-SMC and that a Wright–Fisher evolutionary model suffices in most cases. Using our validated inference framework, we estimate the CNV formation rate at the GAP1 locus in the yeast Saccharomyces cerevisiae to be 10^−4.7 to 10⁻⁴ CNVs per cell division and a fitness coefficient of 0.04 to 0.1 per generation for GAP1 CNVs in glutamine-limited chemostats. We experimentally validated our inference-based estimates using 2 distinct experimental methods—barcode lineage tracking and pairwise fitness assays—which provide independent confirmation of the accuracy of our approach. Our results are consistent with a beneficial CNV supply rate that is 10-fold greater than the estimated rates of beneficial single-nucleotide mutations, explaining the outsized importance of CNVs in rapid adaptive evolution. More generally, our study demonstrates the utility of novel neural network–based likelihood–free inference methods for inferring the rates and effects of evolutionary processes from empirical data with possible applications ranging from tumor to viral evolution.

This study shows that simulation-based inference of evolutionary dynamics using neural networks can yield parameter values for fitness and mutation rate that are difficult to determine experimentally, including those of copy number variants (CNVs) during experimental adaptive evolution of yeast.     

Using protein structural information in evolutionary inference: transmembrane proteins

We present a model of amino acid sequence evolution based on a hidden Markov model that extends to transmembrane proteins previous methods that incorporate protein structural information into phylogenetics. Our model aims to give a better understanding of processes of molecular evolution and to extract structural information from multiple alignments of transmembrane sequences and use such information to improve phylogenetic analyses. This should be of value in phylogenetic studies of transmembrane proteins: for example, mitochondrial proteins have acquired a special importance in phylogenetics and are mostly transmembrane proteins. The improvement in fit to example data sets of our new model relative to less complex models of amino acid sequence evolution is statistically tested. To further illustrate the potential utility of our method, phylogeny estimation is performed on primate CCR5 receptor sequences, sequences of l and m subunits of the light reaction center in purple bacteria, guinea pig sequences with respect to lagomorph and rodent sequences of calcitonin receptor and K-substance receptor, and cetacean sequences of cytochrome b.     

Computational prediction of host-parasite protein interactions between P. falciparum and H. sapiens

To obtain candidates of interactions between proteins of the malaria parasite Plasmodium falciparum and the human host, homologous and conserved interactions were inferred from various sources of interaction data. Such candidate interactions were assessed by applying a machine learning approach and further filtered according to expression and molecular characteristics, enabling involved proteins to indeed interact. The analysis of predicted interactions indicated that parasite proteins predominantly target central proteins to take control of a human host cell. Furthermore, parasite proteins utilized their protein repertoire in a combinatorial manner, providing a broad connection to host cellular processes. In particular, several prominent pathways of signaling and regulation proteins were predicted to interact with parasite chaperones. Such a result suggests an important role of remodeling proteins in the interaction interface between the human host and the parasite. Identification of such molecular strategies that allow the parasite to take control of the host has the potential to deepen our understanding of the parasite specific remodeling processes of the host cell and illuminate new avenues of disease intervention.     

Plasmodium falciparum: selenium-induced cytotoxicity to P. falciparum

The in vitro antimalarial activity of sodium selenite (NaSe) was investigated and the mechanism of its action was studied. NaSe had antimalarial activity against both the chloroquine-susceptible strain FCR-3 and chloroquine-resistant strain K-1 of Plasmodium falciparum. The shrunken cytoplasm of the parasite was observed in a smear 12 h after treatment with NaSe. Co-treatment with copper sulfate (CuSO(4)) in culture did not affect the antimalarial activity of NaSe, but NaSe cytotoxicity against the mammalian cell line Alexander was decreased significantly. The intracellular reduced glutathione level of parasitized red blood cells was decreased significantly by treatment with NaSe, and the decrease was consistent with their mortality. Treatment with NaSe had a strong inhibitory effect on plasmodial development, and NaSe cytotoxicity to human cells was decreased by co-treatment with CuSO(4). These results suggest that co-treatment with NaSe and CuSO(4) may be useful as a new antimalarial therapy.     

Using neural networks and evolutionary information in decoy discrimination for protein tertiary structure prediction

Background  

We present a novel method of protein fold decoy discrimination using machine learning, more specifically using neural networks. Here, decoy discrimination is represented as a machine learning problem, where neural networks are used to learn the native-like features of protein structures using a set of positive and negative training examples. A set of native protein structures provides the positive training examples, while negative training examples are simulated decoy structures obtained by reversing the sequences of native structures. Various features are extracted from the training dataset of positive and negative examples and used as inputs to the neural networks.     

Combining evolutionary information and neural networks to predict protein secondary structure

Using evolutionary information contained in multiple sequence alignments as input to neural networks, secondary structure can be predicted at significantly increased accuracy. Here, we extend our previous three-level system of neural networks by using additional input information derived from multiple alignments. Using a position-specific conservation weight as part of the input increases performance. Using the number of insertions and deletions reduces the tendency for overprediction and increases overall accuracy. Addition of the global amino acid content yields a further improvement, mainly in predicting structural class. The final network system has a sustained overall accuracy of 71.6% in a multiple cross-validation test on 126 unique protein chains. A test on a new set of 124 recently solved protein structures that have no significant sequence similarity to the learning set confirms the high level of accuracy. The average cross-validated accuracy for all 250 sequence-unique chains is above 72%. Using various data sets, the method is compared to alternative prediction methods, some of which also use multiple alignments: the performance advantage of the network system is at least 6 percentage points in three-state accuracy. In addition, the network estimates secondary structure content from multiple sequence alignments about as well as circular dichroism spectroscopy on a single protein and classifies 75% of the 250 proteins correctly into one of four protein structural classes. Of particular practical importance is the definition of a position-specific reliability index. For 40% of all residues the method has a sustained three-state accuracy of 88%, as high as the overall average for homology modelling. A further strength of the method is greatly increased accuracy in predicting the placement of secondary structure segments. © 1994 Wiley-Liss, Inc.     

Using integral projection models to compare population dynamics of four closely related species

Demographic processes, such as survival, growth, and reproduction, can inform us about invasion risk, extinction risk, and trade-offs in life history strategies. The population dynamics of four Amaranthaceae species in southern Illinois, USA were examined using integral projection models (IPMs) to determine whether vital rates reflect life history among these closely related species. Two of the species, Amaranthus palmeri and Amaranthus tuberculatus, are summer annuals and considered to be some of the most problematic agricultural weeds in the US Midwest. Achyranthes japonica is a relatively new invasive exotic species that primarily inhabits forests. Iresine rhizomatosa, is an endangered species in the study area, which also inhabits forests. Two populations of each species were studied from 2012 to 2014 in which height of individuals were measured and used as the state variable in the IPMs. The Amaranthus species and Achyranthes japonica had an estimated population growth rate >1, projecting increases in population size. By contrast, λ was <1 for I. rhizomatosa, projecting a decline in population size demonstrating its endangered status. Germination rates and seed viability were dependent on species and varied over time. Elasticity analyses showed that survival and growth contributed most to λ for the perennial species; whereas, for the annual species population dynamics were driven primarily by fecundity. Overall, Achyranthes japonica and the Amaranthus species show similar trends in demographic processes that align with their invasive nature and not with their life histories. Furthermore, this study demonstrates that more research on the competitive nature of Achyranthes japonica is needed.     

A draft of protein interactions in the malaria parasite P. falciparum

Recent advances have provided a working interactome map for the human malaria parasite Plasmodium falciparum. The aforementioned map, generated from genome-scale analyses, has provided a basis for proteomic studies of the parasite; however, such large-scale approaches commonly suffer from undersampling and lack of coverage. The current map bears no exception, containing only one-quarter of the organism's proteins. Inspired by the needs of the current map and the wealth of bioinformatics data, we assembled a map of 19 979 interactions among 2321 proteins in P. falciparum. The resultant map was generated by computationally inferring protein-protein interactions from evolutionarily conserved protein interactions, underlying domain interactions, and experimental observations. To compile this information into a repository of meaningful data, we assessed interaction quality by applying a logistic regression method, which correlated the presence of an interaction with relevant cellular parameters. Interestingly, it was found that sub-networks from different sources are quite dissimilar in their topologies and overlap to a very small extent. Applying Markov clustering, we observe a typical cluster composition, featuring common cellular functions that were previously reported absent, making this map a valuable resource for understanding the biology of this organism.     

Plasmodium falciparum Plasmodium helical interspersed subtelomeric proteins contribute to cytoadherence and anchor P. falciparum erythrocyte membrane protein 1 to the host cell cytoskeleton

Adherence of Plasmodium falciparum‐infected erythrocytes to host endothelium is conferred through the parasite‐derived virulence factor P. falciparum erythrocyte membrane protein 1 (PfEMP1), the major contributor to malaria severity. PfEMP1 located at knob structures on the erythrocyte surface is anchored to the cytoskeleton, and the Plasmodium helical interspersed subtelomeric (PHIST) gene family plays a role in many host cell modifications including binding the intracellular domain of PfEMP1. Here, we show that conditional reduction of the PHIST protein PFE1605w strongly reduces adhesion of infected erythrocytes to the endothelial receptor CD36. Adhesion to other endothelial receptors was less affected or even unaltered by PFE1605w depletion, suggesting that PHIST proteins might be optimized for subsets of PfEMP1 variants. PFE1605w does not play a role in PfEMP1 transport, but it directly interacts with both the intracellular segment of PfEMP1 and with cytoskeletal components. This is the first report of a PHIST protein interacting with key molecules of the cytoadherence complex and the host cytoskeleton, and this functional role seems to play an essential role in the pathology of P. falciparum.     

Lipopolysaccharide structures of Helicobacter pylori genomic strains 26695 and J99, mouse model H. pylori Sydney strain, H. pylori P466 carrying sialyl Lewis X, and H. pylori UA915 expressing Lewis B classification of H. pylori lipopolysaccharides into glycotype families.

This study describes the molecular makeup of the cell-wall lipopolysaccharides (LPSs) (O-chain polysaccharide-->core oligosaccharide-->lipid A) from five Helicobacter pylori strains: H. pylori 26695 and J99, the complete genome sequences of which have been published, the established mouse model Sydney strain (SS1), and the symptomatic strains P466 and UA915. All chemical and serological experiments were performed on the intact LPSs. H. pylori 26695 and SS1 possessed either a low-Mr semi-rough-form LPS carrying mostly a single Ley type-2 blood-group determinant in the O-chain region covalently attached to the core oligosaccharide or a high-Mr smooth-form LPS, as did strain J99, with an elongated partially fucosylated type-2 N-acetyllactosamine (polyLacNAc) O-chain polymer, terminated mainly by a Lex blood-group determinant, connected to the core oligosaccharide. In the midst of semi-rough-form LPS glycoforms, H. pylori 26695 and SS1 also expressed in the O-chain region a difucosylated antigen, alpha-L-Fucp(1-3)-alpha-L-Fucp(1-4)-beta-D-GlcpNAc, and the cancer-cell-related type-1 or type-2 linear B-blood-group antigen, alpha-D-Galp(1-3)-beta-D-Galp(1-3 or 4)-beta-D-GlcpNAc. The LPS of H. pylori strain P466 carried the cancer-associated type-2 sialyl Lex blood-group antigen, and the LPS from strain UA915 expressed a type-1 Leb blood-group unit. These findings should aid investigations that focus on identifying and characterizing genes responsible for LPS biosynthesis in genomic strains 26695 and J99, and in understanding the role of H. pylori LPS in animal model studies. The LPSs from the H. pylori strains studied to date were grouped into specific glycotype families.     

Evolution and dynamics of protein interactions and networks

The central role of protein-protein interactions (PPIs) in biology has stimulated colossal efforts to identify thousands of them in several organisms. The resulting PPI maps are commonly represented as graphs, where nodes denote proteins and edges represent physical interactions. However, the methods used to generate PPI data on a large scale do not readily allow one to discriminate features such as interaction strength (affinity), type (protein-protein or protein-peptide interaction) or spatiotemporal existence (where and when the proteins are present and interact). Yet, in recent years, a number of studies have tackled these limitations by projecting additional information onto PPIs, revealing novel properties in terms of their evolution and dynamics. In this review we examine these properties both at the binary interaction level and at the network level. We suggest that the diverse and sometimes contradictory results described by different research groups are mostly due to incomplete data coverage and limited data types. Finally, we discuss recently developed methods that will improve this picture in the future.     

The architecture of river networks can drive the evolutionary dynamics of aquatic populations

It is widely recognized that physical landscapes can shape genetic variation within and between populations. However, it is not well understood how riverscapes, with their complex architectures, affect patterns of neutral genetic diversity. Using a spatially explicit agent‐based modeling (ABM) approach, we evaluate the genetic consequences of dendritic river shapes on local population structure. We disentangle the relative contribution of specific river properties to observed patterns of genetic variation by evaluating how different branching architectures and downstream flow regimes affect the genetic structure of populations situated within river networks. Irrespective of the river length, our results illustrate that the extent of river branching, confluence position, and levels of asymmetric downstream migration dictate patterns of genetic variation in riverine populations. Comparisons between simple and highly branched rivers show a 20‐fold increase in the overall genetic diversity and a sevenfold increase in the genetic differentiation between local populations. Given that most rivers have complex architectures, these results highlight the importance of incorporating riverscape information into evolutionary models of aquatic species and could help explain why riverine fishes represent a disproportionately large amount of global vertebrate diversity per unit of habitable area.     

Using evolutionary rates to investigate protein functional divergence and conservation. A case study of the carbonic anhydrases

Functional constraints on proteins limit their evolutionary rates at specific sites. These constraints allow for the interpretation of conserved residues and sites with a rate change as those most likely underlying the functional similarities and differences among protein subfamilies, respectively. This study describes new likelihood-ratio tests (LRTs) that complement existing ones for the identification of both conserved and rate change sites. These identifications are validated by the recovery of residues that are known from existing biochemical and structural information to be critical for the functional similarities and differences among carbonic anhydrases (CAs). In combination with this other information, these LRTs also support a unique antioxidant defense role for the puzzling CA III. As illustrated by the CAs, these LRTs, in combination with other biological evidence, offer a powerful and cost-effective approach for testing hypotheses, making predictions, and designing experiments in protein functional studies.     

Serum antibodies to Helicobacter pylori and its heat-shock protein 60 correlate with the response of gastric mucosa-associated lymphoid tissue lymphoma to eradication of H. pylori

Background and aims. Eradication of Helicobacter pylori leads to regression of mucosa‐associated lymphoid tissue (MALT) lymphomas. In this study, we measured serum antibodies to H. pylori and H. pylori‐recombinant heat‐shock protein 60 (rHSP60) in patients with gastric MALT lymphoma to determine whether humoral immune responses to the bacterial antigens correlate with the efficacy of eradication therapy. Methods. Serum samples were obtained from 33 patients with H. pylori‐positive gastric MALT lymphoma before undergoing therapy to eradicate the bacteria. Anti‐H. pylori antibodies were measured in a commercial assay and in immunoassays to lysates and rHSP60 which were prepared from ATCC 43504 strain. Results. Helicobacter pylori were eradicated in all 33 patients, and the lymphoma completely regressed histologically in 26 patients (79%). Pre‐treatment titers of serum antibody to H. pylori and to rHSP60 in the patients whose tumor regressed were significantly higher than titers in patients whose tumors did not regress (p = .0011 and .035, respectively). By logistic regression analysis, age (odds ratio = 0.88, 95% confidence interval = 0.80–0.99), endoscopic appearance (0.053, 0.004–0.65), titers of anti‐H. pylori antibodies (67.6, 2.5–1800), and titers of anti‐rHSP60 antibody (6.4, 1.2–36) were identified as significantly associated factors with the outcome of MALT lymphoma. Conclusions. Measurement of serum antibodies to H. pylori and HSP60 might be useful for predicting the response of gastric MALT lymphoma to eradication of H. pylori.     

Karyotype comparison between P. chabaudi and P. falciparum: analysis of a P. chabaudi cDNA containing sequences highly repetitive in P. falciparum.

The molecular karyotypes of P. chabaudi and P. falciparum have been compared by pulse field gradient electrophoresis. P. chabaudi has 3 extra chromosomes in the 750-2000 Kb range although the overall number appears to be 14 as is the case for P. falciparum. The chromosomal location of the rRNA genes has been determined for P. chabaudi together with that of a 24 Kd antigen gene. The corresponding cDNA 443 may code for a protein unusually rich in tyrosine and contains sequences highly repetitive in P. falciparum.