Reversible, tunable epigenetic silencing of TCF1 generates flexibility in the T cell memory decision.

The immune system encodes information about the severity of a pathogenic threat in the quantity and type of memory cells it forms. This encoding emerges from lymphocyte decisions to maintain or lose self-renewal and memory potential during a challenge. By tracking CD8+ T cells at the single-cell and clonal lineage level using time-resolved transcriptomics, quantitative live imaging, and an acute infection model, we find that T cells will maintain or lose memory potential early after antigen recognition. However, following pathogen clearance, T cells may regain memory potential if initially lost. Mechanistically, this flexibility is implemented by a stochastic cis-epigenetic switch that tunably and reversibly silences the memory regulator, TCF1, in response to stimulation. Mathematical modeling shows how this flexibility allows memory T cell numbers to scale robustly with pathogen virulence and immune response magnitudes. We propose that flexibility and stochasticity in cellular decisions ensure optimal immune responses against diverse threats.

Steering and controlling evolution - from bioengineering to fighting pathogens.

Control interventions steer the evolution of molecules, viruses, microorganisms or other cells towards a desired outcome. Applications range from engineering biomolecules and synthetic organisms to drug, therapy and vaccine design against pathogens and cancer. In all these instances, a control system alters the eco-evolutionary trajectory of a target system, inducing new functions or suppressing escape evolution. Here, we synthesize the objectives, mechanisms and dynamics of eco-evolutionary control in different biological systems. We discuss how the control system learns and processes information about the target system by sensing or measuring, through adaptive evolution or computational prediction of future trajectories. This information flow distinguishes pre-emptive control strategies by humans from feedback control in biotic systems. We establish a cost-benefit calculus to gauge and optimize control protocols, highlighting the fundamental link between predictability of evolution and efficacy of pre-emptive control.

Learning the shape of protein microenvironments with a holographic convolutional neural network.

Proteins play a central role in biology from immune recognition to brain activity. While major advances in machine learning have improved our ability to predict protein structure from sequence, determining protein function from its sequence or structure remains a major challenge. Here, we introduce holographic convolutional neural network (H-CNN) for proteins, which is a physically motivated machine learning approach to model amino acid preferences in protein structures. H-CNN reflects physical interactions in a protein structure and recapitulates the functional information stored in evolutionary data. H-CNN accurately predicts the impact of mutations on protein stability and binding of protein complexes. Our interpretable computational model for protein structure-function maps could guide design of novel proteins with desired function.

Gene expression variability across cells and species shapes innate immunity.

As the first line of defence against pathogens, cells mount an innate immune response, which varies widely from cell to cell. The response must be potent but carefully controlled to avoid self-damage. How these constraints have shaped the evolution of innate immunity remains poorly understood. Here we characterize the innate immune response's transcriptional divergence between species and variability in expression among cells. Using bulk and single-cell transcriptomics in fibroblasts and mononuclear phagocytes from different species, challenged with immune stimuli, we map the architecture of the innate immune response. Transcriptionally diverging genes, including those that encode cytokines and chemokines, vary across cells and have distinct promoter structures. Conversely, genes that are involved in the regulation of this response, such as those that encode transcription factors and kinases, are conserved between species and display low cell-to-cell variability in expression. We suggest that this expression pattern, which is observed across species and conditions, has evolved as a mechanism for fine-tuned regulation to achieve an effective but balanced response.

Deep generative selection models of T and B cell receptor repertoires with soNNia.

Subclasses of lymphocytes carry different functional roles to work together and produce an immune response and lasting immunity. Additionally to these functional roles, T and B cell lymphocytes rely on the diversity of their receptor chains to recognize different pathogens. The lymphocyte subclasses emerge from common ancestors generated with the same diversity of receptors during selection processes. Here, we leverage biophysical models of receptor generation with machine learning models of selection to identify specific sequence features characteristic of functional lymphocyte repertoires and subrepertoires. Specifically, using only repertoire-level sequence information, we classify CD4+ and CD8+ T cells, find correlations between receptor chains arising during selection, and identify T cell subsets that are targets of pathogenic epitopes. We also show examples of when simple linear classifiers do as well as more complex machine learning methods.

The size of the immune repertoire of bacteria.

Some bacteria and archaea possess an immune system, based on the CRISPR-Cas mechanism, that confers adaptive immunity against viruses. In such species, individual prokaryotes maintain cassettes of viral DNA elements called spacers as a memory of past infections. Typically, the cassettes contain several dozen expressed spacers. Given that bacteria can have very large genomes and since having more spacers should confer a better memory, it is puzzling that so little genetic space would be devoted by prokaryotes to their adaptive immune systems. Here, assuming that CRISPR functions as a long-term memory-based defense against a diverse landscape of viral species, we identify a fundamental tradeoff between the amount of immune memory and effectiveness of response to a given threat. This tradeoff implies an optimal size for the prokaryotic immune repertoire in the observational range.

SOS: online probability estimation and generation of T-and B-cell receptors.

SUMMARY: Recent advances in modelling VDJ recombination and subsequent selection of T- and B-cell receptors provide useful tools to analyse and compare immune repertoires across time, individuals and tissues. A suite of tools-IGoR, OLGA and SONIA-have been publicly released to the community that allow for the inference of generative and selection models from high-throughput sequencing data. However, using these tools requires some scripting or command-line skills and familiarity with complex datasets. As a result, the application of the above models has not been available to a broad audience. In this application note, we fill this gap by presenting Simple OLGA & SONIA (SOS), a web-based interface where users with no coding skills can compute the generation and post-selection probabilities of their sequences, as well as generate batches of synthetic sequences. The application also functions on mobile phones. AVAILABILITY AND IMPLEMENTATION: SOS is freely available to use at sites.google.com/view/statbiophysens/sos with source code at github.com/statbiophys/sos.

Fierce Selection and Interference in B-Cell Repertoire Response to Chronic HIV-1.

During chronic infection, HIV-1 engages in a rapid coevolutionary arms race with the host's adaptive immune system. While it is clear that HIV exerts strong selection on the adaptive immune system, the characteristics of the somatic evolution that shape the immune response are still unknown. Traditional population genetics methods fail to distinguish chronic immune response from healthy repertoire evolution. Here, we infer the evolutionary modes of B-cell repertoires and identify complex dynamics with a constant production of better B-cell receptor (BCR) mutants that compete, maintaining large clonal diversity and potentially slowing down adaptation. A substantial fraction of mutations that rise to high frequencies in pathogen-engaging CDRs of BCRs are beneficial, in contrast to many such changes in structurally relevant frameworks that are deleterious and circulate by hitchhiking. We identify a pattern where BCRs in patients who experience larger viral expansions undergo stronger selection with a rapid turnover of beneficial mutations due to clonal interference in their CDR3 regions. Using population genetics modeling, we show that the extinction of these beneficial mutations can be attributed to the rise of competing beneficial alleles and clonal interference. The picture is of a dynamic repertoire, where better clones may be outcompeted by new mutants before they fix.

Information-Geometric Optimization with Natural Selection.

Evolutionary algorithms, inspired by natural evolution, aim to optimize difficult objective functions without computing derivatives. Here we detail the relationship between classical population genetics of quantitative traits and evolutionary optimization, and formulate a new evolutionary algorithm. Optimization of a continuous objective function is analogous to searching for high fitness phenotypes on a fitness landscape. We describe how natural selection moves a population along the non-Euclidean gradient that is induced by the population on the fitness landscape (the natural gradient). We show how selection is related to Newton's method in optimization under quadratic fitness landscapes, and how selection increases fitness at the cost of reducing diversity. We describe the generation of new phenotypes and introduce an operator that recombines the whole population to generate variants. Finally, we introduce a proof-of-principle algorithm that combines natural selection, our recombination operator, and an adaptive method to increase selection and find the optimum. The algorithm is extremely simple in implementation; it has no matrix inversion or factorization, does not require storing a covariance matrix, and may form the basis of more general model-based optimization algorithms with natural gradient updates.

Host-Pathogen Coevolution and the Emergence of Broadly Neutralizing Antibodies in Chronic Infections.

The vertebrate adaptive immune system provides a flexible and diverse set of molecules to neutralize pathogens. Yet, viruses such as HIV can cause chronic infections by evolving as quickly as the adaptive immune system, forming an evolutionary arms race. Here we introduce a mathematical framework to study the coevolutionary dynamics between antibodies and antigens within a host. We focus on changes in the binding interactions between the antibody and antigen populations, which result from the underlying stochastic evolution of genotype frequencies driven by mutation, selection, and drift. We identify the critical viral and immune parameters that determine the distribution of antibody-antigen binding affinities. We also identify definitive signatures of coevolution that measure the reciprocal response between antibodies and viruses, and we introduce experimentally measurable quantities that quantify the extent of adaptation during continual coevolution of the two opposing populations. Using this analytical framework, we infer rates of viral and immune adaptation based on time-shifted neutralization assays in two HIV-infected patients. Finally, we analyze competition between clonal lineages of antibodies and characterize the fate of a given lineage in terms of the state of the antibody and viral populations. In particular, we derive the conditions that favor the emergence of broadly neutralizing antibodies, which may have relevance to vaccine design against HIV.

Loop-Diffusion: an equivariant diffusion model for designing and scoring protein loops.

Predicting protein functional characteristics from structure remains a central problem in protein science, with broad implications from understanding the mechanisms of disease to designing novel therapeutics. Unfortunately, current machine learning methods are limited by scarce and biased experimental data, and physics-based methods are either too slow to be useful, or too simplified to be accurate. In this work, we present Loop-Diffusion, an energy based diffusion model which leverages a dataset of general protein loops from the entire protein universe to learn an energy function that generalizes to functional prediction tasks. We evaluate Loop-Diffusion's performance on scoring TCR-pMHC interfaces and demonstrate state-of-the-art results in recognizing binding-enhancing mutations.

HERMES: Holographic Equivariant neuRal network model for Mutational Effect and Stability prediction.

Predicting the stability and fitness effects of amino-acid mutations in proteins is a cornerstone of biological discovery and engineering. Various experimental techniques have been developed to measure mutational effects, providing us with extensive datasets across a diverse range of proteins. By training on these data, machine learning approaches have advanced significantly in predicting mutational effects. Here, we introduce HERMES, a 3D rotationally equivariant structure-based neural network model for mutation effect prediction. Pre-trained to predict amino-acid propensities from their surrounding 3D structure atomic environments, HERMES can be efficiently fine-tuned to predict mutational effects, thanks to its symmetry-aware parameterization of the output space. Benchmarking against other models demonstrates that HERMES often outperforms or matches their performance in predicting mutation effects on stability, binding, and fitness, using either computationally or experimentally resolved protein structures. HERMES offers a versatile suit of tools for evaluating mutation effects and can be easily fine-tuned for specific predictive objectives using our open-source code.

HERMES: Holographic Equivariant neuRal network model for Mutational Effect and Stability prediction.

Predicting the stability and fitness effects of amino acid mutations in proteins is a cornerstone of biological discovery and engineering. Various experimental techniques have been developed to measure mutational effects, providing us with extensive datasets across a diverse range of proteins. By training on these data, traditional computational modeling and more recent machine learning approaches have advanced significantly in predicting mutational effects. Here, we introduce HERMES, a 3D rotationally equivariant structure-based neural network model for mutational effect and stability prediction. Pre-trained to predict amino acid propensity from its surrounding 3D structure, HERMES can be fine-tuned for mutational effects using our open-source code. We present a suite of HERMES models, pre-trained with different strategies, and fine-tuned to predict the stability effect of mutations. Benchmarking against other models shows that HERMES often outperforms or matches their performance in predicting mutational effect on stability, binding, and fitness. HERMES offers versatile tools for evaluating mutational effects and can be fine-tuned for specific predictive objectives.

H-Packer: Holographic Rotationally Equivariant Convolutional Neural Network for Protein Side-Chain Packing.

Accurately modeling protein 3D structure is essential for the design of functional proteins. An important sub-task of structure modeling is protein side-chain packing: predicting the conformation of side-chains (rotamers) given the protein's backbone structure and amino-acid sequence. Conventional approaches for this task rely on expensive sampling procedures over hand-crafted energy functions and rotamer libraries. Recently, several deep learning methods have been developed to tackle the problem in a data-driven way, albeit with vastly different formulations (from image-to-image translation to directly predicting atomic coordinates). Here, we frame the problem as a joint regression over the side-chains' true degrees of freedom: the dihedral χ angles. We carefully study possible objective functions for this task, while accounting for the underlying symmetries of the task. We propose Holographic Packer (H-Packer), a novel two-stage algorithm for side-chain packing built on top of two light-weight rotationally equivariant neural networks. We evaluate our method on CASP13 and CASP14 targets. H-Packer is computationally efficient and shows favorable performance against conventional physics-based algorithms and is competitive against alternative deep learning solutions.

Probabilities of developing HIV-1 bNAb sequence features in uninfected and chronically infected individuals.

HIV-1 broadly neutralizing antibodies (bNAbs) are able to suppress viremia and prevent infection. Their induction by vaccination is therefore a major goal. However, in contrast to antibodies that neutralize other pathogens, HIV-1-specific bNAbs frequently carry uncommon molecular characteristics that might prevent their induction. Here, we perform unbiased sequence analyses of B cell receptor repertoires from 57 uninfected and 46 chronically HIV-1- or HCV-infected individuals and learn probabilistic models to predict the likelihood of bNAb development. We formally show that lower probabilities for bNAbs are predictive of higher HIV-1 neutralization activity. Moreover, ranking bNAbs by their probabilities allows to identify highly potent antibodies with superior generation probabilities as preferential targets for vaccination approaches. Importantly, we find equal bNAb probabilities across infected and uninfected individuals. This implies that chronic infection is not a prerequisite for the generation of bNAbs, fostering the hope that HIV-1 vaccines can induce bNAb development in uninfected people.

Formation of regulatory modules by local sequence duplication.

Turnover of regulatory sequence and function is an important part of molecular evolution. But what are the modes of sequence evolution leading to rapid formation and loss of regulatory sites? Here we show that a large fraction of neighboring transcription factor binding sites in the fly genome have formed from a common sequence origin by local duplications. This mode of evolution is found to produce regulatory information: duplications can seed new sites in the neighborhood of existing sites. Duplicate seeds evolve subsequently by point mutations, often towards binding a different factor than their ancestral neighbor sites. These results are based on a statistical analysis of 346 cis-regulatory modules in the Drosophila melanogaster genome, and a comparison set of intergenic regulatory sequence in Saccharomyces cerevisiae. In fly regulatory modules, pairs of binding sites show significantly enhanced sequence similarity up to distances of about 50 bp. We analyze these data in terms of an evolutionary model with two distinct modes of site formation: (i) evolution from independent sequence origin and (ii) divergent evolution following duplication of a common ancestor sequence. Our results suggest that pervasive formation of binding sites by local sequence duplications distinguishes the complex regulatory architecture of higher eukaryotes from the simpler architecture of unicellular organisms.

T cell immune responses deciphered.

A machine-learning approach reveals antigen encoding that predicts T cell responses.

Mutual information maximization for amortized likelihood inference from sampled trajectories: MINIMALIST.

Simulation-based inference enables learning the parameters of a model even when its likelihood cannot be computed in practice. One class of methods uses data simulated with different parameters to infer models of the likelihood-to-evidence ratio, or equivalently the posterior function. Here we frame the inference task as an estimation of an energy function parametrized with an artificial neural network. We present an intuitive approach, named MINIMALIST, in which the optimal model of the likelihood-to-evidence ratio is found by maximizing the likelihood of simulated data. Within this framework, the connection between the task of simulation-based inference and mutual information maximization is clear, and we show how several known methods of posterior estimation relate to alternative lower bounds to mutual information. These distinct objective functions aim at the same optimal energy form and therefore can be directly benchmarked. We compare their accuracy in the inference of model parameters, focusing on four dynamical systems that encompass common challenges in time series analysis: dynamics driven by multiplicative noise, nonlinear interactions, chaotic behavior, and high-dimensional parameter space.

Design of an optimal combination therapy with broadly neutralizing antibodies to suppress HIV-1.

Infusion of broadly neutralizing antibodies (bNAbs) has shown promise as an alternative to anti-retroviral therapy against HIV. A key challenge is to suppress viral escape, which is more effectively achieved with a combination of bNAbs. Here, we propose a computational approach to predict the efficacy of a bNAb therapy based on the population genetics of HIV escape, which we parametrize using high-throughput HIV sequence data from bNAb-naive patients. By quantifying the mutational target size and the fitness cost of HIV-1 escape from bNAbs, we predict the distribution of rebound times in three clinical trials. We show that a cocktail of three bNAbs is necessary to effectively suppress viral escape, and predict the optimal composition of such bNAb cocktail. Our results offer a rational therapy design for HIV, and show how genetic data can be used to predict treatment outcomes and design new approaches to pathogenic control.

Optimal evolutionary decision-making to store immune memory.

The adaptive immune system provides a diverse set of molecules that can mount specific responses against a multitude of pathogens. Memory is a key feature of adaptive immunity, which allows organisms to respond more readily upon re-infections. However, differentiation of memory cells is still one of the least understood cell fate decisions. Here, we introduce a mathematical framework to characterize optimal strategies to store memory to maximize the utility of immune response over an organism's lifetime. We show that memory production should be actively regulated to balance between affinity and cross-reactivity of immune receptors for an effective protection against evolving pathogens. Moreover, we predict that specificity of memory should depend on the organism's lifespan, and shorter lived organisms with fewer pathogenic encounters should store more cross-reactive memory. Our framework provides a baseline to gauge the efficacy of immune memory in light of an organism's coevolutionary history with pathogens.