Structural and mechanistic basis of σ-dependent transcriptional pausing.

In σ-dependent transcriptional pausing, the transcription initiation factor σ, translocating with RNA polymerase (RNAP), makes sequence-specific protein­DNA interactions with a promoter-like sequence element in the transcribed region, inducing pausing. It has been proposed that, in σ-dependent pausing, the RNAP active center can access off-pathway "backtracked" states that are substrates for the transcript-cleavage factors of the Gre family and on-pathway "scrunched" states that mediate pause escape. Here, using site-specific protein­DNA photocrosslinking to define positions of the RNAP trailing and leading edges and of σ relative to DNA at the λPR' promoter, we show directly that σ-dependent pausing in the absence of GreB in vitro predominantly involves a state backtracked by 2­4 bp, and σ-dependent pausing in the presence of GreB in vitro and in vivo predominantly involves a state scrunched by 2­3 bp. Analogous experiments with a library of 47 (∼16,000) transcribed-region sequences show that the state scrunched by 2­3 bp­and only that state­is associated with the consensus sequence, T−3N−2Y−1G+1, (where −1 corresponds to the position of the RNA 3' end), which is identical to the consensus for pausing in initial transcription and which is related to the consensus for pausing in transcription elongation. Experiments with heteroduplex templates show that sequence information at position T−3 resides in the DNA nontemplate strand. A cryoelectron microscopy structure of a complex engaged in σ-dependent pausing reveals positions of DNA scrunching on the DNA nontemplate and template strands and suggests that position T−3 of the consensus sequence exerts its effects by facilitating scrunching.

Novel tool to quantify with single-cell resolution the number of incoming AAV genomes co-expressed in the mouse nervous system.

Adeno-associated viral (AAV) vectors are an established and safe gene delivery tool to target the nervous system. However, the payload capacity of <4.9 kb limits the transfer of large or multiple genes. Oversized payloads could be delivered by fragmenting the transgenes into separate AAV capsids that are then mixed. This strategy could increase the AAV cargo capacity to treat monogenic, polygenic diseases and comorbidities only if controlled co-expression of multiple AAV capsids is achieved on each transduced cell. We developed a tool to quantify the number of incoming AAV genomes that are co-expressed in the nervous system with single-cell resolution. By using an isogenic mix of three AAVs each expressing single fluorescent reporters, we determined that expression of much greater than 31 AAV genomes per neuron in vitro and 20 genomes per neuron in vivo is obtained across different brain regions including anterior cingulate, prefrontal, somatomotor and somatosensory cortex areas, and cerebellar lobule VI. Our results demonstrate that multiple AAV vectors containing different transgenes or transgene fragments, can efficiently co-express in the same neuron. This tool can be used to design and improve AAV-based interrogation of neuronal circuits, map brain connectivity, and treat genetic diseases affecting the nervous system.

Manifold learning for organizing unstructured sets of process observations.

Data mining is routinely used to organize ensembles of short temporal observations so as to reconstruct useful, low-dimensional realizations of an underlying dynamical system. In this paper, we use manifold learning to organize unstructured ensembles of observations ("trials") of a system's response surface. We have no control over where every trial starts, and during each trial, operating conditions are varied by turning "agnostic" knobs, which change system parameters in a systematic, but unknown way. As one (or more) knobs "turn," we record (possibly partial) observations of the system response. We demonstrate how such partial and disorganized observation ensembles can be integrated into coherent response surfaces whose dimension and parametrization can be systematically recovered in a data-driven fashion. The approach can be justified through the Whitney and Takens embedding theorems, allowing reconstruction of manifolds/attractors through different types of observations. We demonstrate our approach by organizing unstructured observations of response surfaces, including the reconstruction of a cusp bifurcation surface for hydrogen combustion in a continuous stirred tank reactor. Finally, we demonstrate how this observation-based reconstruction naturally leads to informative transport maps between the input parameter space and output/state variable spaces.

Coarse-scale PDEs from fine-scale observations via machine learning.

Complex spatiotemporal dynamics of physicochemical processes are often modeled at a microscopic level (through, e.g., atomistic, agent-based, or lattice models) based on first principles. Some of these processes can also be successfully modeled at the macroscopic level using, e.g., partial differential equations (PDEs) describing the evolution of the right few macroscopic observables (e.g., concentration and momentum fields). Deriving good macroscopic descriptions (the so-called "closure problem") is often a time-consuming process requiring deep understanding/intuition about the system of interest. Recent developments in data science provide alternative ways to effectively extract/learn accurate macroscopic descriptions approximating the underlying microscopic observations. In this paper, we introduce a data-driven framework for the identification of unavailable coarse-scale PDEs from microscopic observations via machine-learning algorithms. Specifically, using Gaussian processes, artificial neural networks, and/or diffusion maps, the proposed framework uncovers the relation between the relevant macroscopic space fields and their time evolution (the right-hand side of the explicitly unavailable macroscopic PDE). Interestingly, several choices equally representative of the data can be discovered. The framework will be illustrated through the data-driven discovery of macroscopic, concentration-level PDEs resulting from a fine-scale, lattice Boltzmann level model of a reaction/transport process. Once the coarse evolution law is identified, it can be simulated to produce long-term macroscopic predictions. Different features (pros as well as cons) of alternative machine-learning algorithms for performing this task (Gaussian processes and artificial neural networks) are presented and discussed.

Spectral Quasi-Equilibrium Manifold for Chemical Kinetics.

The Spectral Quasi-Equilibrium Manifold (SQEM) method is a model reduction technique for chemical kinetics based on entropy maximization under constraints built by the slowest eigenvectors at equilibrium. The method is revisited here and discussed and validated through the Michaelis-Menten kinetic scheme, and the quality of the reduction is related to the temporal evolution and the gap between eigenvalues. SQEM is then applied to detailed reaction mechanisms for the homogeneous combustion of hydrogen, syngas, and methane mixtures with air in adiabatic constant pressure reactors. The system states computed using SQEM are compared with those obtained by direct integration of the detailed mechanism, and good agreement between the reduced and the detailed descriptions is demonstrated. The SQEM reduced model of hydrogen/air combustion is also compared with another similar technique, the Rate-Controlled Constrained-Equilibrium (RCCE). For the same number of representative variables, SQEM is found to provide a more accurate description.

The global relaxation redistribution method for reduction of combustion kinetics.

An algorithm based on the Relaxation Redistribution Method (RRM) is proposed for constructing the Slow Invariant Manifold (SIM) of a chosen dimension to cover a large fraction of the admissible composition space that includes the equilibrium and initial states. The manifold boundaries are determined with the help of the Rate Controlled Constrained Equilibrium method, which also provides the initial guess for the SIM. The latter is iteratively refined until convergence and the converged manifold is tabulated. A criterion based on the departure from invariance is proposed to find the region over which the reduced description is valid. The global realization of the RRM algorithm is applied to constant pressure auto-ignition and adiabatic premixed laminar flames of hydrogen-air mixtures.

Specificity, synergy, and mechanisms of splice-modifying drugs.

Drugs that target pre-mRNA splicing hold great therapeutic potential, but the quantitative understanding of how these drugs work is limited. Here we introduce mechanistically interpretable quantitative models for the sequence-specific and concentration-dependent behavior of splice-modifying drugs. Using massively parallel splicing assays, RNA-seq experiments, and precision dose-response curves, we obtain quantitative models for two small-molecule drugs, risdiplam and branaplam, developed for treating spinal muscular atrophy. The results quantitatively characterize the specificities of risdiplam and branaplam for 5' splice site sequences, suggest that branaplam recognizes 5' splice sites via two distinct interaction modes, and contradict the prevailing two-site hypothesis for risdiplam activity at SMN2 exon 7. The results also show that anomalous single-drug cooperativity, as well as multi-drug synergy, are widespread among small-molecule drugs and antisense-oligonucleotide drugs that promote exon inclusion. Our quantitative models thus clarify the mechanisms of existing treatments and provide a basis for the rational development of new therapies.

Machine-learning-based data-driven discovery of nonlinear phase-field dynamics.

One of the main questions regarding complex systems at large scales concerns the effective interactions and driving forces that emerge from the detailed microscopic properties. Coarse-grained models aim to describe complex systems in terms of coarse-scale equations with a reduced number of degrees of freedom. Recent developments in machine-learning algorithms have significantly empowered the discovery process of governing equations directly from data. However, it remains difficult to discover partial differential equations (PDEs) with high-order derivatives. In this paper, we present data-driven architectures based on a multilayer perceptron, a convolutional neural network (CNN), and a combination of a CNN and long short-term memory structures for discovering the nonlinear equations of motion for phase-field models with nonconserved and conserved order parameters. The well-known Allen-Cahn, Cahn-Hilliard, and phase-field crystal models were used as test cases. Two conceptually different types of implementations were used: (a) guided by physical intuition (such as the local dependence of the derivatives) and (b) in the absence of any physical assumptions (black-box model). We show that not only can we effectively learn the time derivatives of the field in both scenarios, but we can also use the data-driven PDEs to propagate the field in time and achieve results in good agreement with the original PDEs.

MAVE-NN: learning genotype-phenotype maps from multiplex assays of variant effect.

Multiplex assays of variant effect (MAVEs) are a family of methods that includes deep mutational scanning experiments on proteins and massively parallel reporter assays on gene regulatory sequences. Despite their increasing popularity, a general strategy for inferring quantitative models of genotype-phenotype maps from MAVE data is lacking. Here we introduce MAVE-NN, a neural-network-based Python package that implements a broadly applicable information-theoretic framework for learning genotype-phenotype maps-including biophysically interpretable models-from MAVE datasets. We demonstrate MAVE-NN in multiple biological contexts, and highlight the ability of our approach to deconvolve mutational effects from otherwise confounding experimental nonlinearities and noise.

On the parameter combinations that matter and on those that do not: data-driven studies of parameter (non)identifiability.

We present a data-driven approach to characterizing nonidentifiability of a model's parameters and illustrate it through dynamic as well as steady kinetic models. By employing Diffusion Maps and their extensions, we discover the minimal combinations of parameters required to characterize the output behavior of a chemical system: a set of effective parameters for the model. Furthermore, we introduce and use a Conformal Autoencoder Neural Network technique, as well as a kernel-based Jointly Smooth Function technique, to disentangle the redundant parameter combinations that do not affect the output behavior from the ones that do. We discuss the interpretability of our data-driven effective parameters, and demonstrate the utility of the approach both for behavior prediction and parameter estimation. In the latter task, it becomes important to describe level sets in parameter space that are consistent with a particular output behavior. We validate our approach on a model of multisite phosphorylation, where a reduced set of effective parameters (nonlinear combinations of the physical ones) has previously been established analytically.

Emergent Spaces for Coupled Oscillators.

Systems of coupled dynamical units (e.g., oscillators or neurons) are known to exhibit complex, emergent behaviors that may be simplified through coarse-graining: a process in which one discovers coarse variables and derives equations for their evolution. Such coarse-graining procedures often require extensive experience and/or a deep understanding of the system dynamics. In this paper we present a systematic, data-driven approach to discovering "bespoke" coarse variables based on manifold learning algorithms. We illustrate this methodology with the classic Kuramoto phase oscillator model, and demonstrate how our manifold learning technique can successfully identify a coarse variable that is one-to-one with the established Kuramoto order parameter. We then introduce an extension of our coarse-graining methodology which enables us to learn evolution equations for the discovered coarse variables via an artificial neural network architecture templated on numerical time integrators (initial value solvers). This approach allows us to learn accurate approximations of time derivatives of state variables from sparse flow data, and hence discover useful approximate differential equation descriptions of their dynamic behavior. We demonstrate this capability by learning ODEs that agree with the known analytical expression for the Kuramoto order parameter dynamics at the continuum limit. We then show how this approach can also be used to learn the dynamics of coarse variables discovered through our manifold learning methodology. In both of these examples, we compare the results of our neural network based method to typical finite differences complemented with geometric harmonics. Finally, we present a series of computational examples illustrating how a variation of our manifold learning methodology can be used to discover sets of "effective" parameters, reduced parameter combinations, for multi-parameter models with complex coupling. We conclude with a discussion of possible extensions of this approach, including the possibility of obtaining data-driven effective partial differential equations for coarse-grained neuronal network behavior, as illustrated by the synchronization dynamics of Hodgkin-Huxley type neurons with a Chung-Lu network. Thus, we build an integrated suite of tools for obtaining data-driven coarse variables, data-driven effective parameters, and data-driven coarse-grained equations from detailed observations of networks of oscillators.

Manifold learning for parameter reduction.

Large scale dynamical systems (e.g. many nonlinear coupled differential equations) can often be summarized in terms of only a few state variables (a few equations), a trait that reduces complexity and facilitates exploration of behavioral aspects of otherwise intractable models. High model dimensionality and complexity makes symbolic, pen-and-paper model reduction tedious and impractical, a difficulty addressed by recently developed frameworks that computerize reduction. Symbolic work has the benefit, however, of identifying both reduced state variables and parameter combinations that matter most (effective parameters, "inputs"); whereas current computational reduction schemes leave the parameter reduction aspect mostly unaddressed. As the interest in mapping out and optimizing complex input-output relations keeps growing, it becomes clear that combating the curse of dimensionality also requires efficient schemes for input space exploration and reduction. Here, we explore systematic, data-driven parameter reduction by means of effective parameter identification, starting from current nonlinear manifoldlearning techniques enabling state space reduction. Our approach aspires to extend the data-driven determination of effective state variables with the data-driven discovery of effective model parameters, and thus to accelerate the exploration of high-dimensional parameter spaces associated with complex models.