Chemical genomic analysis reveals the interplay between iron chelation, zinc homeostasis and retromer function in the bioactivity of an ethanol adduct of the feijoa fruit-derived ellagitannin vescalagin.

Nature has been a rich source of pharmaceutical compounds, producing 80% of our currently prescribed drugs. The feijoa plant, Acca sellowiana, is classified in the family Myrtaceae, native to South America, and currently grown worldwide to produce feijoa fruit. Feijoa is a rich source of bioactive compounds with anticancer, anti-inflammatory, antibacterial and antifungal activities; however, the mechanism of action of these compounds are largely not known. Here we used chemical genetic analyses in the model organism Saccharomyces cerevisiae to investigate the mechanism of action of a feijoa-derived ethanol adduct of vescalagin (EtOH-vescalagin). Genome-wide barcode sequencing (Bar-seq) analysis revealed yeast strains lacking genes in iron metabolism, zinc metabolism, retromer function or mitochondrial function were hypersensitive to 0.3 µM EtOH-vescalagin. This treatment increased expression of iron uptake proteins at the plasma membrane, which was a compensatory response to reduced intracellular iron. Likewise, EtOH-vescalagin increased expression of the Cot1 protein in the vacuolar membrane that transports zinc into the vacuole to prevent cytoplasmic accumulation of zinc. Each individual subunit in the retromer complex was required for the iron homeostatic mechanism of EtOH-vescalagin, while only the cargo recognition component in the retromer complex was required for the zinc homeostatic mechanism. Overexpression of either retromer subunits or high-affinity iron transporters suppressed EtOH-vescalagin bioactivity in a zinc-replete condition, while overexpression of only retromer subunits increased EtOH-vescalagin bioactivity in a zinc-deficient condition. Together, these results indicate that EtOH-vescalagin bioactivity begins with extracellular iron chelation and proceeds with intracellular transport of zinc via the retromer complex. More broadly, this is the first report of a bioactive compound to further characterize the poorly understood interaction between zinc metabolism and retromer function.

Host obesity impacts genetic variation in influenza A viral populations.

Obesity is well established as a risk factor for many noncommunicable diseases; however, its consequences for infectious disease are poorly understood. Here, we investigated the impact of host obesity on influenza A virus (IAV) genetic variation using a diet-induced obesity ferret model and the A/Hong Kong/1073/1999 (H9N2) strain. Using a co-caging study design, we investigated the maintenance, generation, and transmission of intrahost IAV genetic variation by sequencing viral genomic RNA obtained from nasal wash samples over multiple days of infection. We found evidence for an enhanced role of positive selection acting on de novo mutations in obese hosts that led to nonsynonymous changes that rose to high frequency. In addition, we identified numerous cases of mutations throughout the genome that were specific to obese hosts and that were preserved during transmission between hosts. Despite detection of obese-specific variants, the overall viral genetic diversity did not differ significantly between obese and lean hosts. This is likely due to the high supply rate of de novo variation and common evolutionary adaptations to the ferret host regardless of obesity status, which we show are mediated by variation in the hemagglutinin and polymerase genes (PB2 and PB1). We also identified defective viral genomes (DVGs) that were found uniquely in either obese or lean hosts, but the overall DVG diversity and dynamics did not differ between the two groups. Our study suggests that obesity may result in a unique selective environment impacting intrahost IAV evolution, highlighting the need for additional genetic and functional studies to confirm these effects.IMPORTANCEObesity is a chronic health condition characterized by excess adiposity leading to a systemic increase in inflammation and dysregulation of metabolic hormones and immune cell populations. Influenza A virus (IAV) is a highly infectious pathogen responsible for seasonal and pandemic influenza. Host risk factors, including compromised immunity and pre-existing health conditions, can contribute to increased infection susceptibility and disease severity. During viral replication in a host, the negative-sense single-stranded RNA genome of IAV accumulates genetic diversity that may have important consequences for viral evolution and transmission. Our study provides the first insight into the consequences of host obesity on viral genetic diversity and adaptation, suggesting that host factors associated with obesity alter the selective environment experienced by a viral population, thereby impacting the spectrum of genetic variation.

Patient Subtyping Analysis of Baseline Multi-omic Data Reveals Distinct Pre-immune States Predictive of Vaccination Responses.

Understanding the molecular mechanisms that underpin diverse vaccination responses is a critical step toward developing efficient vaccines. Molecular subtyping approaches can offer valuable insights into the heterogeneous nature of responses and aid in the design of more effective vaccines. In order to explore the molecular signatures associated with the vaccine response, we analyzed baseline transcriptomics data from paired samples of whole blood, proteomics and glycomics data from serum, and metabolomics data from urine, obtained from influenza vaccine recipients (2019-2020 season) prior to vaccination. After integrating the data using a network-based model, we performed a subtyping analysis. The integration of multiple data modalities from 62 samples resulted in five baseline molecular subtypes with distinct molecular signatures. These baseline subtypes differed in the expression of pre-existing adaptive or innate immunity signatures, which were linked to significant variation across subtypes in baseline immunoglobulin A (IgA) and hemagglutination inhibition (HAI) titer levels. It is worth noting that these significant differences persisted through day 28 post-vaccination, indicating the effect of initial immune state on vaccination response. These findings highlight the significance of interpersonal variation in baseline immune status as a crucial factor in determining vaccine response and efficacy. Ultimately, incorporating molecular profiling could enable personalized vaccine optimization.

Structure-primed embedding on the transcription factor manifold enables transparent model architectures for gene regulatory network and latent activity inference.

BACKGROUND: Modeling of gene regulatory networks (GRNs) is limited due to a lack of direct measurements of genome-wide transcription factor activity (TFA) making it difficult to separate covariance and regulatory interactions. Inference of regulatory interactions and TFA requires aggregation of complementary evidence. Estimating TFA explicitly is problematic as it disconnects GRN inference and TFA estimation and is unable to account for, for example, contextual transcription factor-transcription factor interactions, and other higher order features. Deep-learning offers a potential solution, as it can model complex interactions and higher-order latent features, although does not provide interpretable models and latent features. RESULTS: We propose a novel autoencoder-based framework, StrUcture Primed Inference of Regulation using latent Factor ACTivity (SupirFactor) for modeling, and a metric, explained relative variance (ERV), for interpretation of GRNs. We evaluate SupirFactor with ERV in a wide set of contexts. Compared to current state-of-the-art GRN inference methods, SupirFactor performs favorably. We evaluate latent feature activity as an estimate of TFA and biological function in S. cerevisiae as well as in peripheral blood mononuclear cells (PBMC). CONCLUSION: Here we present a framework for structure-primed inference and interpretation of GRNs, SupirFactor, demonstrating interpretability using ERV in multiple biological and experimental settings. SupirFactor enables TFA estimation and pathway analysis using latent factor activity, demonstrated here on two large-scale single-cell datasets, modeling S. cerevisiae and PBMC. We find that the SupirFactor model facilitates biological analysis acquiring novel functional and regulatory insight.

Post-transcriptional mechanisms modulate the consequences of adaptive copy number variation.

Copy-number variants (CNVs) are large-scale amplifications or deletions of DNA that can drive rapid adaptive evolution and result in large-scale changes in gene expression. Whereas alterations in the copy number of one or more genes within a CNV can confer a selective advantage, other genes within a CNV can decrease fitness when their dosage is changed. Dosage compensation - in which the gene expression output from multiple gene copies is less than expected - is one means by which an organism can mitigate the fitness costs of deleterious gene amplification. Previous research has shown evidence for dosage compensation at both the transcriptional level and at the level of protein expression; however, the extent of compensation differs substantially between genes, strains, and studies. Here, we investigated sources of dosage compensation at multiple levels of gene expression regulation by defining the transcriptome, translatome and proteome of experimentally evolved yeast (Saccharomyces cerevisiae) strains containing adaptive CNVs. We quantified the gene expression output at each step and found evidence of widespread dosage compensation at the protein abundance (~47%) level. By contrast we find only limited evidence for dosage compensation at the transcriptional (~8%) and translational (~3%) level. We also find substantial divergence in the expression of unamplified genes in evolved strains that could be due to either the presence of a CNV or adaptation to the environment. Detailed analysis of 82 amplified and 411 unamplified genes with significantly discrepant relationships between RNA and protein abundances identified enrichment for upstream open reading frames (uORFs). These uORFs are enriched for binding site motifs for SSD1, an RNA binding protein that has previously been associated with tolerance of aneuploidy. Our findings suggest that, in the presence of CNVs, SSD1 may act to alter the expression of specific genes by potentiating uORF mediated translational regulation.

Simultaneous estimation of gene regulatory network structure and RNA kinetics from single cell gene expression.

Cells respond to environmental and developmental stimuli by remodeling their transcriptomes through regulation of both mRNA transcription and mRNA decay. A central goal of biology is identifying the global set of regulatory relationships between factors that control mRNA production and degradation and their target transcripts and construct a predictive model of gene expression. Regulatory relationships are typically identified using transcriptome measurements and causal inference algorithms. RNA kinetic parameters are determined experimentally by employing run-on or metabolic labeling (e.g. 4-thiouracil) methods that allow transcription and decay rates to be separately measured. Here, we develop a deep learning model, trained with single-cell RNA-seq data, that both infers causal regulatory relationships and estimates RNA kinetic parameters. The resulting in silico model predicts future gene expression states and can be perturbed to simulate the effect of transcription factor changes. We acquired model training data by sequencing the transcriptomes of 175,000 individual Saccharomyces cerevisiae cells that were subject to an external perturbation and continuously sampled over a one hour period. The rate of change for each transcript was calculated on a per-cell basis to estimate RNA velocity. We then trained a deep learning model with transcriptome and RNA velocity data to calculate time-dependent estimates of mRNA production and decay rates. By separating RNA velocity into transcription and decay rates, we show that rapamycin treatment causes existing ribosomal protein transcripts to be rapidly destabilized, while production of new transcripts gradually slows over the course of an hour. The neural network framework we present is designed to explicitly model causal regulatory relationships between transcription factors and their genes, and shows superior performance to existing models on the basis of recovery of known regulatory relationships. We validated the predictive power of the model by perturbing transcription factors in silico and comparing transcriptome-wide effects with experimental data. Our study represents the first step in constructing a complete, predictive, biophysical model of gene expression regulation.

Copy number variation alters local and global mutational tolerance.

Copy number variants (CNVs), duplications and deletions of genomic sequences, contribute to evolutionary adaptation but can also confer deleterious effects and cause disease. Whereas the effects of amplifying individual genes or whole chromosomes (i.e., aneuploidy) have been studied extensively, much less is known about the genetic and functional effects of CNVs of differing sizes and structures. Here, we investigated Saccharomyces cerevisiae (yeast) strains that acquired adaptive CNVs of variable structures and copy numbers following experimental evolution in glutamine-limited chemostats. Although beneficial in the selective environment, CNVs result in decreased fitness compared with the euploid ancestor in rich media. We used transposon mutagenesis to investigate mutational tolerance and genome-wide genetic interactions in CNV strains. We find that CNVs increase mutational target size, confer increased mutational tolerance in amplified essential genes, and result in novel genetic interactions with unlinked genes. We validated a novel genetic interaction between different CNVs and BMH1 that was common to multiple strains. We also analyzed global gene expression and found that transcriptional dosage compensation does not affect most genes amplified by CNVs, although gene-specific transcriptional dosage compensation does occur for ∼12% of amplified genes. Furthermore, we find that CNV strains do not show previously described transcriptional signatures of aneuploidy. Our study reveals the extent to which local and global mutational tolerance is modified by CNVs with implications for genome evolution and CNV-associated diseases, such as cancer.

Parallel proteomics and phosphoproteomics defines starvation signal specific processes in cell quiescence.

Cells arrest growth and enter a quiescent state upon nutrient deprivation. However, the molecular processes by which cells respond to different starvation signals to regulate exit from the cell division cycle and initiation of quiescence remains poorly understood. To study the role of protein expression and signaling in quiescence we combined temporal profiling of the proteome and phosphoproteome using stable isotope labeling with amino acids in cell culture (SILAC) in Saccharomyces cerevisiae (budding yeast). We find that carbon and phosphorus starvation signals activate quiescence through largely distinct remodeling of the proteome and phosphoproteome. However, increased expression of mitochondrial proteins is associated with quiescence establishment in response to both starvation signals. Deletion of the putative quiescence regulator RIM15, which encodes a serine-threonine kinase, results in reduced survival of cells starved for phosphorus and nitrogen, but not carbon. However, we identified common protein phosphorylation roles for RIM15 in quiescence that are enriched for RNA metabolism and translation. We also find evidence for RIM15-mediated phosphorylation of some targets, including IGO1, prior to starvation consistent with a functional role for RIM15 in proliferative cells. Finally, our results reveal widespread catabolism of amino acids in response to nitrogen starvation, indicating widespread amino acid recycling via salvage pathways in conditions lacking environmental nitrogen. Our study defines an expanded quiescent proteome and phosphoproteome in yeast, and highlights the multiple coordinated molecular processes at the level of protein expression and phosphorylation that are required for quiescence.

mRNA condensation fluidizes the cytoplasm.

The intracellular environment is packed with macromolecules of mesoscale size, and this crowded milieu significantly influences cell physiology. When exposed to stress, mRNAs released after translational arrest condense with RNA binding proteins, resulting in the formation of membraneless RNA protein (RNP) condensates known as processing bodies (P-bodies) and stress granules (SGs). However, the impact of the assembly of these condensates on the biophysical properties of the crowded cytoplasmic environment remains unclear. Here, we find that upon exposure to stress, polysome collapse and condensation of mRNAs increases mesoscale particle diffusivity in the cytoplasm. Increased mesoscale diffusivity is required for the efficient formation of Q-bodies, membraneless organelles that coordinate degradation of misfolded peptides that accumulate during stress. Additionally, we demonstrate that polysome collapse and stress granule formation has a similar effect in mammalian cells, fluidizing the cytoplasm at the mesoscale. We find that synthetic, light-induced RNA condensation is sufficient to fluidize the cytoplasm, demonstrating a causal effect of RNA condensation. Together, our work reveals a new functional role for stress-induced translation inhibition and formation of RNP condensates in modulating the physical properties of the cytoplasm to effectively respond to stressful conditions.

Host obesity impacts genetic variation in influenza A viral populations.

Obesity is a chronic health condition characterized by excess adiposity leading to a systemic increase in inflammation and dysregulation of metabolic hormones and immune cell populations. Obesity is well established as a risk factor for many noncommunicable diseases; however, its consequences for infectious disease are poorly understood. Influenza A virus (IAV) is a highly infectious pathogen responsible for seasonal and pandemic influenza. Host risk factors, including compromised immunity and pre-existing health conditions, can contribute to increased infection susceptibility and disease severity. During viral replication in a host, the negative sense single stranded RNA genome of IAV accumulates genetic diversity that may have important consequences for viral evolution and transmission. Here, we investigated the impact of host obesity on IAV genetic variation using a diet induced obesity ferret model. We infected obese and lean male ferrets with the A/Hong Kong/1073/1999 (H9N2) IAV strain. Using a co-caging study design, we investigated the maintenance, generation, and transmission of intrahost IAV genetic variation by sequencing viral genomic RNA obtained from nasal wash samples over multiple days of infection. We found evidence for an enhanced role of positive selection acting on de novo mutations in obese hosts that led to nonsynonymous changes that rose to high frequency. In addition, we identified numerous cases of recurrent low-frequency mutations throughout the genome that were specific to obese hosts. Despite these obese-specific variants, overall viral genetic diversity did not differ significantly between obese and lean hosts. This is likely due to the high supply rate of de novo variation and common evolutionary adaptations to the ferret host regardless of obesity status, which we show are mediated by variation in the hemagglutinin (HA) and polymerase genes (PB2 and PB1). As with single nucleotide variants, we identified a class of defective viral genomes (DVGs) that were found uniquely in either obese or lean hosts, but overall DVG diversity and dynamics did not differ between the two groups. Our study provides the first insight into the consequences of host obesity on viral genetic diversity and adaptation, suggesting that host factors associated with obesity alter the selective environment experienced by a viral population, thereby impacting the spectrum of genetic variation.

Best Practices in Microbial Experimental Evolution: Using Reporters and Long-Read Sequencing to Identify Copy Number Variation in Experimental Evolution.

Copy number variants (CNVs), comprising gene amplifications and deletions, are a pervasive class of heritable variation. CNVs play a key role in rapid adaptation in both natural, and experimental, evolution. However, despite the advent of new DNA sequencing technologies, detection and quantification of CNVs in heterogeneous populations has remained challenging. Here, we summarize recent advances in the use of CNV reporters that provide a facile means of quantifying de novo CNVs at a specific locus in the genome, and nanopore sequencing, for resolving the often complex structures of CNVs. We provide guidance for the engineering and analysis of CNV reporters and practical guidelines for single-cell analysis of CNVs using flow cytometry. We summarize recent advances in nanopore sequencing, discuss the utility of this technology, and provide guidance for the bioinformatic analysis of these data to define the molecular structure of CNVs. The combination of reporter systems for tracking and isolating CNV lineages and long-read DNA sequencing for characterizing CNV structures enables unprecedented resolution of the mechanisms by which CNVs are generated and their evolutionary dynamics.

Influenza A virus reassortment is strain dependent.

RNA viruses can exchange genetic material during coinfection, an interaction that creates novel strains with implications for viral evolution and public health. Influenza A viral genetic exchange can occur when genome segments from distinct strains reassort in coinfected cells. Predicting potential genomic reassortment between influenza strains has been a long-standing goal. Experimental coinfection studies have shed light on factors that limit or promote reassortment. However, determining the reassortment potential between diverse Influenza A strains has remained elusive. To address this challenge, we developed a high throughput genotyping approach to quantify reassortment among a diverse panel of human influenza virus strains encompassing two pandemics (swine and avian origin), three specific epidemics, and both circulating human subtypes A/H1N1 and A/H3N2. We found that reassortment frequency (the proportion of reassortants generated) is an emergent property of specific pairs of strains where strain identity is a predictor of reassortment frequency. We detect little evidence that antigenic subtype drives reassortment as intersubtype (H1N1xH3N2) and intrasubtype reassortment frequencies were, on average, similar. Instead, our data suggest that certain strains bias the reassortment frequency up or down, independently of the coinfecting partner. We observe that viral productivity is also an emergent property of coinfections, but uncorrelated to reassortment frequency; thus viral productivity is a separate factor affecting the total number of reassortants produced. Assortment of individual segments among progeny and pairwise segment combinations within progeny generally favored homologous combinations. These outcomes were not related to strain similarity or shared subtype but reassortment frequency was closely correlated to the proportion of both unique genotypes and of progeny with heterologous pairwise segment combinations. We provide experimental evidence that viral genetic exchange is potentially an individual social trait subject to natural selection, which implies the propensity for reassortment is not evenly shared among strains. This study highlights the need for research incorporating diverse strains to discover the traits that shift the reassortment potential to realize the goal of predicting influenza virus evolution resulting from segment exchange.

Increased mesoscale diffusivity in response to acute glucose starvation.

Macromolecular crowding is an important property of cells that impacts multiple biological processes. Passive microrheology using single particle tracking is a powerful means of studying macromolecular crowding. Here we monitored the diffusivity of self-assembling fluorescent nanoparticles (µNS) and mRNPs ( GFA1 -PP7) in response to acute glucose starvation. mRNP diffusivity was reduced upon glucose starvation as previously reported. By contrast, we observed increased diffusivity of µNS particles. Our results suggest that, upon glucose starvation, mRNP granule diffusivity may be reduced due to increased physical interactions, whereas macromolecular crowding in the cytoplasm may be globally reduced.

Structure primed embedding on the transcription factor manifold enables transparent model architectures for gene regulatory network and latent activity inference.

The modeling of gene regulatory networks (GRNs) is limited due to a lack of direct measurements of regulatory features in genome-wide screens. Most GRN inference methods are therefore forced to model relationships between regulatory genes and their targets with expression as a proxy for the upstream independent features, complicating validation and predictions produced by modeling frameworks. Separating covariance and regulatory influence requires aggregation of independent and complementary sets of evidence, such as transcription factor (TF) binding and target gene expression. However, the complete regulatory state of the system, e.g. TF activity (TFA) is unknown due to a lack of experimental feasibility, making regulatory relations difficult to infer. Some methods attempt to account for this by modeling TFA as a latent feature, but these models often use linear frameworks that are unable to account for non-linearities such as saturation, TF-TF interactions, and other higher order features. Deep learning frameworks may offer a solution, as they are capable of modeling complex interactions and capturing higher-order latent features. However, these methods often discard central concepts in biological systems modeling, such as sparsity and latent feature interpretability, in favor of increased model complexity. We propose a novel deep learning autoencoder-based framework, StrUcture Primed Inference of Regulation using latent Factor ACTivity (SupirFactor), that scales to single cell genomic data and maintains interpretability to perform GRN inference and estimate TFA as a latent feature. We demonstrate that SupirFactor outperforms current leading GRN inference methods, predicts biologically relevant TFA and elucidates functional regulatory pathways through aggregation of TFs.

Metagenomes from Soils along an Agricultural Transect in Ulster County, New York.

Many modern farming practices negatively impact ecosystems on the local and global scales. Here, we assessed the taxonomic structures of 48 soil microbial communities along an agricultural transect using 16S rRNA and internal transcribed spacer (ITS) amplicon sequencing. We further characterized the functional structures of a subsample of 12 microbiomes using whole-genome sequencing.

Increased mesoscale diffusivity in response to acute glucose starvation.

Macromolecular crowding is an important parameter that impacts multiple biological processes. Passive microrheology using single particle tracking is a powerful means of studying macromolecular crowding. Here we monitored the diffusivity of self-assembling fluorescent nanoparticles (µNS) in response to acute glucose starvation. mRNP diffusivity was reduced upon glucose starvation as previously reported. In contrast, we observed increased diffusivity of µNS particles. Our results suggest that, upon glucose starvation, mRNP granule diffusivity may be reduced due to changes in physical interactions, while global crowding in the cytoplasm may be reduced.

Vaccination History, Body Mass Index, Age, and Baseline Gene Expression Predict Influenza Vaccination Outcomes.

Seasonal influenza is a primary public health burden in the USA and globally. Annual vaccination programs are designed on the basis of circulating influenza viral strains. However, the effectiveness of the seasonal influenza vaccine is highly variable between seasons and among individuals. A number of factors are known to influence vaccination effectiveness including age, sex, and comorbidities. Here, we sought to determine whether whole blood gene expression profiling prior to vaccination is informative about pre-existing immunological status and the immunological response to vaccine. We performed whole transcriptome analysis using RNA sequencing (RNAseq) of whole blood samples obtained prior to vaccination from 275 participants enrolled in an annual influenza vaccine trial. Serological status prior to vaccination and 28 days following vaccination was assessed using the hemagglutination inhibition assay (HAI) to define baseline immune status and the response to vaccination. We find evidence that genes with immunological functions are increased in expression in individuals with higher pre-existing immunity and in those individuals who mount a greater response to vaccination. Using a random forest model, we find that this set of genes can be used to predict vaccine response with a performance similar to a model that incorporates physiological and prior vaccination status alone. A model using both gene expression and physiological factors has the greatest predictive power demonstrating the potential utility of molecular profiling for enhancing prediction of vaccine response. Moreover, expression of genes that are associated with enhanced vaccination response may point to additional biological pathways that contribute to mounting a robust immunological response to the seasonal influenza vaccine.

Complex Genomic Rearrangements following Selection in a Glutamine-Limited Medium over Hundreds of Generations.

Large-scale genomic changes, including copy number variations (CNVs), are frequently observed in long-term evolution experiments (LTEEs). We have previously reported the detection of recurrent CNVs in Saccharomyces cerevisiae populations adapting to glutamine-limited conditions over hundreds of generations. Here, we present the whole-genome sequencing (WGS) assemblies of 7 LTEE strains and their ancestor.

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.

High-performance single-cell gene regulatory network inference at scale: the Inferelator 3.0.

MOTIVATION: Gene regulatory networks define regulatory relationships between transcription factors and target genes within a biological system, and reconstructing them is essential for understanding cellular growth and function. Methods for inferring and reconstructing networks from genomics data have evolved rapidly over the last decade in response to advances in sequencing technology and machine learning. The scale of data collection has increased dramatically; the largest genome-wide gene expression datasets have grown from thousands of measurements to millions of single cells, and new technologies are on the horizon to increase to tens of millions of cells and above. RESULTS: In this work, we present the Inferelator 3.0, which has been significantly updated to integrate data from distinct cell types to learn context-specific regulatory networks and aggregate them into a shared regulatory network, while retaining the functionality of the previous versions. The Inferelator is able to integrate the largest single-cell datasets and learn cell-type-specific gene regulatory networks. Compared to other network inference methods, the Inferelator learns new and informative Saccharomyces cerevisiae networks from single-cell gene expression data, measured by recovery of a known gold standard. We demonstrate its scaling capabilities by learning networks for multiple distinct neuronal and glial cell types in the developing Mus musculus brain at E18 from a large (1.3 million) single-cell gene expression dataset with paired single-cell chromatin accessibility data. AVAILABILITY AND IMPLEMENTATION: The inferelator software is available on GitHub (https://github.com/flatironinstitute/inferelator) under the MIT license and has been released as python packages with associated documentation (https://inferelator.readthedocs.io/). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.