Enhancing molecular design efficiency: Uniting language models and generative networks with genetic algorithms.

This study examines the effectiveness of generative models in drug discovery, material science, and polymer science, aiming to overcome constraints associated with traditional inverse design methods relying on heuristic rules. Generative models generate synthetic data resembling real data, enabling deep learning model training without extensive labeled datasets. They prove valuable in creating virtual libraries of molecules for material science and facilitating drug discovery by generating molecules with specific properties. While generative adversarial networks (GANs) are explored for these purposes, mode collapse restricts their efficacy, limiting novel structure variability. To address this, we introduce a masked language model (LM) inspired by natural language processing. Although LMs alone can have inherent limitations, we propose a hybrid architecture combining LMs and GANs to efficiently generate new molecules, demonstrating superior performance over standalone masked LMs, particularly for smaller population sizes. This hybrid LM-GAN architecture enhances efficiency in optimizing properties and generating novel samples.

Approaching Maximum Resolution in Structured Illumination Microscopy via Accurate Noise Modeling.

Biological images captured by microscopes are characterized by heterogeneous signal-to-noise ratios (SNRs) due to spatially varying photon emission across the field of view convoluted with camera noise. State-of-the-art unsupervised structured illumination microscopy (SIM) reconstruction algorithms, commonly implemented in the Fourier domain, do not accurately model this noise and suffer from high-frequency artifacts, user-dependent choices of smoothness constraints making assumptions on biological features, and unphysical negative values in the recovered fluorescence intensity map. On the other hand, supervised methods rely on large datasets for training, and often require retraining for new sample structures. Consequently, achieving high contrast near the maximum theoretical resolution in an unsupervised, physically principled, manner remains an open problem. Here, we propose Bayesian-SIM (B-SIM), an unsupervised Bayesian framework to quantitatively reconstruct SIM data, rectifying these shortcomings by accurately incorporating known noise sources in the spatial domain. To accelerate the reconstruction process, we use the finite extent of the point-spread-function to devise a parallelized Monte Carlo strategy involving chunking and restitching of the inferred fluorescence intensity. We benchmark our framework on both simulated and experimental images, and demonstrate improved contrast permitting feature recovery at up to 25% shorter length scales over state-of-the-art methods at both high- and low-SNR. B-SIM enables unsupervised, quantitative, physically accurate reconstruction without the need for labeled training data, democratizing high-quality SIM reconstruction and expands the capabilities of live-cell SIM to lower SNR, potentially revealing biological features in previously inaccessible regimes.

Transferring a Molecular Foundation Model for Polymer Property Predictions.

Transformer-based large language models have remarkable potential to accelerate design optimization for applications such as drug development and material discovery. Self-supervised pretraining of transformer models requires large-scale data sets, which are often sparsely populated in topical areas such as polymer science. State-of-the-art approaches for polymers conduct data augmentation to generate additional samples but unavoidably incur extra computational costs. In contrast, large-scale open-source data sets are available for small molecules and provide a potential solution to data scarcity through transfer learning. In this work, we show that using transformers pretrained on small molecules and fine-tuned on polymer properties achieves comparable accuracy to those trained on augmented polymer data sets for a series of benchmark prediction tasks.

Using mechanistic models and machine learning to design single-color multiplexed nascent chain tracking experiments.

mRNA translation is the ubiquitous cellular process of reading messenger-RNA strands into functional proteins. Over the past decade, large strides in microscopy techniques have allowed observation of mRNA translation at a single-molecule resolution for self-consistent time-series measurements in live cells. Dubbed Nascent chain tracking (NCT), these methods have explored many temporal dynamics in mRNA translation uncaptured by other experimental methods such as ribosomal profiling, smFISH, pSILAC, BONCAT, or FUNCAT-PLA. However, NCT is currently restricted to the observation of one or two mRNA species at a time due to limits in the number of resolvable fluorescent tags. In this work, we propose a hybrid computational pipeline, where detailed mechanistic simulations produce realistic NCT videos, and machine learning is used to assess potential experimental designs for their ability to resolve multiple mRNA species using a single fluorescent color for all species. Our simulation results show that with careful application this hybrid design strategy could in principle be used to extend the number of mRNA species that could be watched simultaneously within the same cell. We present a simulated example NCT experiment with seven different mRNA species within the same simulated cell and use our ML labeling to identify these spots with 90% accuracy using only two distinct fluorescent tags. We conclude that the proposed extension to the NCT color palette should allow experimentalists to access a plethora of new experimental design possibilities, especially for cell Signaling applications requiring simultaneous study of multiple mRNAs.

Bayesian filtering for model predictive control of stochastic gene expression in single cells.

This study describes a method for controlling the production of protein in individual cells using stochastic models of gene expression. By combining modern microscopy platforms with optogenetic gene expression, experimentalists are able to accurately apply light to individual cells, which can induce protein production. Here we use a finite state projection based stochastic model of gene expression, along with Bayesian state estimation to control protein copy numbers within individual cells. We compare this method to previous methods that use population based approaches. We also demonstrate the ability of this control strategy to ameliorate discrepancies between the predictions of a deterministic model and stochastic switching system.

Adaptive language model training for molecular design.

The vast size of chemical space necessitates computational approaches to automate and accelerate the design of molecular sequences to guide experimental efforts for drug discovery. Genetic algorithms provide a useful framework to incrementally generate molecules by applying mutations to known chemical structures. Recently, masked language models have been applied to automate the mutation process by leveraging large compound libraries to learn commonly occurring chemical sequences (i.e., using tokenization) and predict rearrangements (i.e., using mask prediction). Here, we consider how language models can be adapted to improve molecule generation for different optimization tasks. We use two different generation strategies for comparison, fixed and adaptive. The fixed strategy uses a pre-trained model to generate mutations; the adaptive strategy trains the language model on each new generation of molecules selected for target properties during optimization. Our results show that the adaptive strategy allows the language model to more closely fit the distribution of molecules in the population. Therefore, for enhanced fitness optimization, we suggest the use of the fixed strategy during an initial phase followed by the use of the adaptive strategy. We demonstrate the impact of adaptive training by searching for molecules that optimize both heuristic metrics, drug-likeness and synthesizability, as well as predicted protein binding affinity from a surrogate model. Our results show that the adaptive strategy provides a significant improvement in fitness optimization compared to the fixed pre-trained model, empowering the application of language models to molecular design tasks.

Using Mechanistic Models and Machine Learning to Design Single-Color Multiplexed Nascent Chain Tracking Experiments.

mRNA translation is the ubiquitous cellular process of reading messenger-RNA strands into functional proteins. Over the past decade, large strides in microscopy techniques have allowed observation of mRNA translation at a single-molecule resolution for self-consistent time-series measurements in live cells. Dubbed Nascent chain tracking (NCT), these methods have explored many temporal dynamics in mRNA translation uncaptured by other experimental methods such as ribosomal profiling, smFISH, pSILAC, BONCAT, or FUNCAT-PLA. However, NCT is currently restricted to the observation of one or two mRNA species at a time due to limits in the number of resolvable fluorescent tags. In this work, we propose a hybrid computational pipeline, where detailed mechanistic simulations produce realistic NCT videos, and machine learning is used to assess potential experimental designs for their ability to resolve multiple mRNA species using a single fluorescent color for all species. Through simulation, we show that with careful application, this hybrid design strategy could in principle be used to extend the number of mRNA species that could be watched simultaneously within the same cell. We present a simulated example NCT experiment with seven different mRNA species within the same simulated cell and use our ML labeling to identify these spots with 90% accuracy using only two distinct fluorescent tags. The proposed extension to the NCT color palette should allow experimentalists to access a plethora of new experimental design possibilities, especially for cell signalling applications requiring simultaneous study of multiple mRNAs.

Enabling reactive microscopy with MicroMator.

Microscopy image analysis has recently made enormous progress both in terms of accuracy and speed thanks to machine learning methods and improved computational resources. This greatly facilitates the online adaptation of microscopy experimental plans using real-time information of the observed systems and their environments. Applications in which reactiveness is needed are multifarious. Here we report MicroMator, an open and flexible software for defining and driving reactive microscopy experiments. It provides a Python software environment and an extensible set of modules that greatly facilitate the definition of events with triggers and effects interacting with the experiment. We provide a pedagogic example performing dynamic adaptation of fluorescence illumination on bacteria, and demonstrate MicroMator's potential via two challenging case studies in yeast to single-cell control and single-cell recombination, both requiring real-time tracking and light targeting at the single-cell level.

Femtosecond X-ray Spectroscopy Directly Quantifies Transient Excited-State Mixed Valency.

Quantifying charge delocalization associated with short-lived photoexcited states of molecular complexes in solution remains experimentally challenging, requiring local element specific femtosecond experimental probes of time-evolving electron transfer. In this study, we quantify the evolving valence hole charge distribution in the photoexcited charge transfer state of a prototypical mixed valence bimetallic iron-ruthenium complex, [(CN)5FeIICNRuIII(NH3)5]-, in water by combining femtosecond X-ray spectroscopy measurements with time-dependent density functional theory calculations of the excited-state dynamics. We estimate the valence hole charge that accumulated at the Fe atom to be 0.6 ± 0.2, resulting from excited-state metal-to-metal charge transfer, on an ∼60 fs time scale. Our combined experimental and computational approach provides a spectroscopic ruler for quantifying excited-state valency in solvated complexes.

Aging power spectrum of membrane protein transport and other subordinated random walks.

Single-particle tracking offers detailed information about the motion of molecules in complex environments such as those encountered in live cells, but the interpretation of experimental data is challenging. One of the most powerful tools in the characterization of random processes is the power spectral density. However, because anomalous diffusion processes in complex systems are usually not stationary, the traditional Wiener-Khinchin theorem for the analysis of power spectral densities is invalid. Here, we employ a recently developed tool named aging Wiener-Khinchin theorem to derive the power spectral density of fractional Brownian motion coexisting with a scale-free continuous time random walk, the two most typical anomalous diffusion processes. Using this analysis, we characterize the motion of voltage-gated sodium channels on the surface of hippocampal neurons. Our results show aging where the power spectral density can either increase or decrease with observation time depending on the specific parameters of both underlying processes.

Building predictive signaling models by perturbing yeast cells with time-varying stimulations resulting in distinct signaling responses.

This protocol provides a step-by-step approach to perturb single cells with time-varying stimulation profiles, collect distinct signaling responses, and use these to infer a system of ordinary differential equations to capture and predict dynamics of protein-protein regulation in signal transduction pathways. The models are validated by predicting the signaling activation upon new cell stimulation conditions. In comparison to using standard step-like stimulations, application of diverse time-varying cell stimulations results in better inference of model parameters and substantially improves model predictions. For complete details on the use and results of this protocol, please refer to Jashnsaz et al. (2020).

Pleuropulmonary Blastoma in Pediatric Lung Lesions.

BACKGROUND: Pediatric lung lesions are a group of mostly benign pulmonary anomalies with a broad spectrum of clinical disease and histopathology. Our objective was to evaluate the characteristics of children undergoing resection of a primary lung lesion and to identify preoperative risk factors for malignancy. METHODS: A retrospective cohort study was conducted by using an operative database of 521 primary lung lesions managed at 11 children's hospitals in the United States. Multivariable logistic regression was used to examine the relationship between preoperative characteristics and risk of malignancy, including pleuropulmonary blastoma (PPB). RESULTS: None of the 344 prenatally diagnosed lesions had malignant pathology (P < .0001). Among 177 children without a history of prenatal detection, 15 (8.7%) were classified as having a malignant tumor (type 1 PPB, n = 11; other PPB, n = 3; adenocarcinoma, n = 1) at a median age of 20.7 months (interquartile range, 7.9-58.1). Malignancy was associated with the DICER1 mutation in 8 (57%) PPB cases. No malignant lesion had a systemic feeding vessel (P = .0427). The sensitivity of preoperative chest computed tomography (CT) for detecting malignant pathology was 33.3% (95% confidence interval [CI]: 15.2-58.3). Multivariable logistic regression revealed that increased suspicion of malignancy by CT and bilateral disease were significant predictors of malignant pathology (odds ratios of 42.15 [95% CI, 7.43-340.3; P < .0001] and 42.03 [95% CI, 3.51-995.6; P = .0041], respectively). CONCLUSIONS: In pediatric lung masses initially diagnosed after birth, the risk of PPB approached 10%. These results strongly caution against routine nonoperative management in this patient population. DICER1 testing may be helpful given the poor sensitivity of CT for identifying malignant pathology.

Direct observation of coherent femtosecond solvent reorganization coupled to intramolecular electron transfer.

It is well known that the solvent plays a critical role in ultrafast electron-transfer reactions. However, solvent reorganization occurs on multiple length scales, and selectively measuring short-range solute-solvent interactions at the atomic level with femtosecond time resolution remains a challenge. Here we report femtosecond X-ray scattering and emission measurements following photoinduced charge-transfer excitation in a mixed-valence bimetallic (FeiiRuiii) complex in water, and their interpretation using non-equilibrium molecular dynamics simulations. Combined experimental and computational analysis reveals that the charge-transfer excited state has a lifetime of 62 fs and that coherent translational motions of the first solvation shell are coupled to the back electron transfer. Our molecular dynamics simulations identify that the observed coherent translational motions arise from hydrogen bonding changes between the solute and nearby water molecules upon photoexcitation, and have an amplitude of tenths of ångströms, 120-200 cm-1 frequency and ~100 fs relaxation time. This study provides an atomistic view of coherent solvent reorganization mediating ultrafast intramolecular electron transfer.

Diverse Cell Stimulation Kinetics Identify Predictive Signal Transduction Models.

Computationally understanding the molecular mechanisms that give rise to cell signaling responses upon different environmental, chemical, and genetic perturbations is a long-standing challenge that requires models that fit and predict quantitative responses for new biological conditions. Overcoming this challenge depends not only on good models and detailed experimental data but also on the rigorous integration of both. We propose a quantitative framework to perturb and model generic signaling networks using multiple and diverse changing environments (hereafter "kinetic stimulations") resulting in distinct pathway activation dynamics. We demonstrate that utilizing multiple diverse kinetic stimulations better constrains model parameters and enables predictions of signaling dynamics that would be impossible using traditional dose-response or individual kinetic stimulations. To demonstrate our approach, we use experimentally identified models to predict signaling dynamics in normal, mutated, and drug-treated conditions upon multitudes of kinetic stimulations and quantify which proteins and reaction rates are most sensitive to which extracellular stimulations.

Optimal Design of Single-Cell Experiments within Temporally Fluctuating Environments.

Modern biological experiments are becoming increasingly complex, and designing these experiments to yield the greatest possible quantitative insight is an open challenge. Increasingly, computational models of complex stochastic biological systems are being used to understand and predict biological behaviors or to infer biological parameters. Such quantitative analyses can also help to improve experiment designs for particular goals, such as to learn more about specific model mechanisms or to reduce prediction errors in certain situations. A classic approach to experiment design is to use the Fisher information matrix (FIM), which quantifies the expected information a particular experiment will reveal about model parameters. The Finite State Projection based FIM (FSP-FIM) was recently developed to compute the FIM for discrete stochastic gene regulatory systems, whose complex response distributions do not satisfy standard assumptions of Gaussian variations. In this work, we develop the FSP-FIM analysis for a stochastic model of stress response genes in S. cerevisae under time-varying MAPK induction. We verify this FSP-FIM analysis and use it to optimize the number of cells that should be quantified at particular times to learn as much as possible about the model parameters. We then extend the FSP-FIM approach to explore how different measurement times or genetic modifications help to minimize uncertainty in the sensing of extracellular environments, and we experimentally validate the FSP-FIM to rank single-cell experiments for their abilities to minimize estimation uncertainty of NaCl concentrations during yeast osmotic shock. This work demonstrates the potential of quantitative models to not only make sense of modern biological data sets, but to close the loop between quantitative modeling and experimental data collection.

SPI-Hub™: a gateway to scholarly publishing information.

BACKGROUND: Advances in the health sciences rely on sharing research and data through publication. As information professionals are often asked to contribute their knowledge to assist clinicians and researchers in selecting journals for publication, the authors recognized an opportunity to build a decision support tool, SPI-Hub: Scholarly Publishing Information Hub™, to capture the team's collective publishing industry knowledge, while carefully retaining the quality of service. CASE PRESENTATION: SPI-Hub's decision support functionality relies on a data framework that describes journal publication policies and practices through a newly designed metadata structure, the Knowledge Management Journal Record™. Metadata fields are populated through a semi-automated process that uses custom programming to access content from multiple sources. Each record includes 25 metadata fields representing best publishing practices. Currently, the database includes more than 24,000 health sciences journal records. To correctly capture the resources needed for both completion and future maintenance of the project, the team conducted an internal study to assess time requirements for completing records through different stages of automation. CONCLUSIONS: The journal decision support tool, SPI-Hub, provides an opportunity to assess publication practices by compiling data from a variety of sources in a single location. Automated and semi-automated approaches have effectively reduced the time needed for data collection. Through a comprehensive knowledge management framework and the incorporation of multiple quality points specific to each journal, SPI-Hub provides prospective users with both recommendations for publication and holistic assessment of the trustworthiness of journals in which to publish research and acquire trusted knowledge.

Determining the Orientation and Vibronic Couplings between Electronic and Vibrational Coordinates with Polarization-Selective Two-Dimensional Vibrational-Electronic Spectroscopy.

We experimentally demonstrate polarization-selective two-dimensional (2D) vibrational-electronic (VE) spectroscopy on a transition-metal mixed-valence complex where the cyanide stretching vibrations are coupled to the metal-to-metal charge-transfer transition. A simultaneous fitting of the parallel and crossed polarized 2D VE spectra quantifies the relative vibronic coupling strengths and angles between the charge-transfer transition and three coupled cyanide stretching vibrations in a mode-specific manner. In particular, we find that the bridging vibration, which modulates the distance between the transition-metal centers, is oriented nearly parallel to the charge-transfer axis and is 9 times more strongly coupled to the electronic transition than the radial vibration, which is oriented almost perpendicular to the charge-transfer axis. The results from this experiment allow us to map the spectroscopically observed vibronic coordinates onto the molecular frame providing a general method to spatially resolve vibronic energy transfer on a femtosecond time scale.

Why equating all evidence searches to systematic reviews defies their role in information seeking.

All too often the quality and rigor of topic investigations is inaccurately conveyed to information professionals, resulting in a mischaracterization of the research, which, if left unchecked and published, may in turn mislead potential readers. Accurately understanding and categorizing the types of topic investigation searches that are requested of information professionals is critical to both meeting requestors' needs and reflecting their intended methodological approaches. Information professionals' expertise can be an invaluable resource to guide users through the investigative and publication process.

Computational design and interpretation of single-RNA translation experiments.

Advances in fluorescence microscopy have introduced new assays to quantify live-cell translation dynamics at single-RNA resolution. We introduce a detailed, yet efficient sequence-based stochastic model that generates realistic synthetic data for several such assays, including Fluorescence Correlation Spectroscopy (FCS), ribosome Run-Off Assays (ROA) after Harringtonine application, and Fluorescence Recovery After Photobleaching (FRAP). We simulate these experiments under multiple imaging conditions and for thousands of human genes, and we evaluate through simulations which experiments are most likely to provide accurate estimates of elongation kinetics. Finding that FCS analyses are optimal for both short and long length genes, we integrate our model with experimental FCS data to capture the nascent protein statistics and temporal dynamics for three human genes: KDM5B, ß-actin, and H2B. Finally, we introduce a new open-source software package, RNA Sequence to NAscent Protein Simulator (rSNAPsim), to easily simulate the single-molecule translation dynamics of any gene sequence for any of these assays and for different assumptions regarding synonymous codon usage, tRNA level modifications, or ribosome pauses. rSNAPsim is implemented in Python and is available at: https://github.com/MunskyGroup/rSNAPsim.git.