Cell Type Purification by Single-Cell Transcriptome-Trained Sorting.

Much of current molecular and cell biology research relies on the ability to purify cell types by fluorescence-activated cell sorting (FACS). FACS typically relies on the ability to label cell types of interest with antibodies or fluorescent transgenic constructs. However, antibody availability is often limited, and genetic manipulation is labor intensive or impossible in the case of primary human tissue. To date, no systematic method exists to enrich for cell types without a priori knowledge of cell-type markers. Here, we propose GateID, a computational method that combines single-cell transcriptomics with FACS index sorting to purify cell types of choice using only native cellular properties such as cell size, granularity, and mitochondrial content. We validate GateID by purifying various cell types from zebrafish kidney marrow and the human pancreas to high purity without resorting to specific antibodies or transgenes.

Protein Evolution and Design.

This article introduces the Protein Evolution and Design theme of the Annual Review of Biochemistry Volume 87.

Machine-learning-optimized Cas12a barcoding enables the recovery of single-cell lineages and transcriptional profiles.

The development of CRISPR-based barcoding methods creates an exciting opportunity to understand cellular phylogenies. We present a compact, tunable, high-capacity Cas12a barcoding system called dual acting inverted site array (DAISY). We combined high-throughput screening and machine learning to predict and optimize the 60-bp DAISY barcode sequences. After optimization, top-performing barcodes had ∼10-fold increased capacity relative to the best random-screened designs and performed reliably across diverse cell types. DAISY barcode arrays generated ∼12 bits of entropy and ∼66,000 unique barcodes. Thus, DAISY barcodes-at a fraction of the size of Cas9 barcodes-achieved high-capacity barcoding. We coupled DAISY barcoding with single-cell RNA-seq to recover lineages and gene expression profiles from ∼47,000 human melanoma cells. A single DAISY barcode recovered up to ∼700 lineages from one parental cell. This analysis revealed heritable single-cell gene expression and potential epigenetic modulation of memory gene transcription. Overall, Cas12a DAISY barcoding is an efficient tool for investigating cell-state dynamics.

Artificial Intelligence for Drug Discovery: Are We There Yet?

Drug discovery is adapting to novel technologies such as data science, informatics, and artificial intelligence (AI) to accelerate effective treatment development while reducing costs and animal experiments. AI is transforming drug discovery, as indicated by increasing interest from investors, industrial and academic scientists, and legislators. Successful drug discovery requires optimizing properties related to pharmacodynamics, pharmacokinetics, and clinical outcomes. This review discusses the use of AI in the three pillars of drug discovery: diseases, targets, and therapeutic modalities, with a focus on small-molecule drugs. AI technologies, such as generative chemistry, machine learning, and multiproperty optimization, have enabled several compounds to enter clinical trials. The scientific community must carefully vet known information to address the reproducibility crisis. The full potential of AI in drug discovery can only be realized with sufficient ground truth and appropriate human intervention at later pipeline stages.

Mechanistic regulation of planarian shape during growth and degrowth.

Adult planarians can grow when fed and degrow (shrink) when starved while maintaining their whole-body shape. It is unknown how the morphogens patterning the planarian axes are coordinated during feeding and starvation or how they modulate the necessary differential tissue growth or degrowth. Here, we investigate the dynamics of planarian shape together with a theoretical study of the mechanisms regulating whole-body proportions and shape. We found that the planarian body proportions scale isometrically following similar linear rates during growth and degrowth, but that fed worms are significantly wider than starved worms. By combining a descriptive model of planarian shape and size with a mechanistic model of anterior-posterior and medio-lateral signaling calibrated with a novel parameter optimization methodology, we theoretically demonstrate that the feedback loop between these positional information signals and the shape they control can regulate the planarian whole-body shape during growth. Furthermore, the computational model produced the correct shape and size dynamics during degrowth as a result of a predicted increase in apoptosis rate and pole signal during starvation. These results offer mechanistic insights into the dynamic regulation of whole-body morphologies.

Quantization avoids saddle points in distributed optimization.

Distributed nonconvex optimization underpins key functionalities of numerous distributed systems, ranging from power systems, smart buildings, cooperative robots, vehicle networks to sensor networks. Recently, it has also merged as a promising solution to handle the enormous growth in data and model sizes in deep learning. A fundamental problem in distributed nonconvex optimization is avoiding convergence to saddle points, which significantly degrade optimization accuracy. We find that the process of quantization, which is necessary for all digital communications, can be exploited to enable saddle-point avoidance. More specifically, we propose a stochastic quantization scheme and prove that it can effectively escape saddle points and ensure convergence to a second-order stationary point in distributed nonconvex optimization. With an easily adjustable quantization granularity, the approach allows a user to control the number of bits sent per iteration and, hence, to aggressively reduce the communication overhead. Numerical experimental results using distributed optimization and learning problems on benchmark datasets confirm the effectiveness of the approach.

Discovering optimal kinetic pathways for self-assembly using automatic differentiation.

Macromolecular complexes are often composed of diverse subunits. The self-assembly of these subunits is inherently nonequilibrium and must avoid kinetic traps to achieve high yield over feasible timescales. We show how the kinetics of self-assembly benefits from diversity in subunits because it generates an expansive parameter space that naturally improves the "expressivity" of self-assembly, much like a deeper neural network. By using automatic differentiation algorithms commonly used in deep learning, we searched the parameter spaces of mass-action kinetic models to identify classes of kinetic protocols that mimic biological solutions for productive self-assembly. Our results reveal how high-yield complexes that easily become kinetically trapped in incomplete intermediates can instead be steered by internal design of rate-constants or external and active control of subunits to efficiently assemble. Internal design of a hierarchy of subunit binding rates generates self-assembly that can robustly avoid kinetic traps for all concentrations and energetics, but it places strict constraints on selection of relative rates. External control via subunit titration is more versatile, avoiding kinetic traps for any system without requiring molecular engineering of binding rates, albeit less efficiently and robustly. We derive theoretical expressions for the timescales of kinetic traps, and we demonstrate our optimization method applies not just for design but inference, extracting intersubunit binding rates from observations of yield-vs.-time for a heterotetramer. Overall, we identify optimal kinetic protocols for self-assembly as a powerful mechanism to achieve efficient and high-yield assembly in synthetic systems whether robustness or ease of "designability" is preferred.

The training process of many deep networks explores the same low-dimensional manifold.

We develop information-geometric techniques to analyze the trajectories of the predictions of deep networks during training. By examining the underlying high-dimensional probabilistic models, we reveal that the training process explores an effectively low-dimensional manifold. Networks with a wide range of architectures, sizes, trained using different optimization methods, regularization techniques, data augmentation techniques, and weight initializations lie on the same manifold in the prediction space. We study the details of this manifold to find that networks with different architectures follow distinguishable trajectories, but other factors have a minimal influence; larger networks train along a similar manifold as that of smaller networks, just faster; and networks initialized at very different parts of the prediction space converge to the solution along a similar manifold.

Simulating institutional heterogeneity in sustainability science.

Sustainability outcomes are influenced by the laws and configurations of natural and engineered systems as well as activities in socio-economic systems. An important subset of human activity is the creation and implementation of institutions, formal and informal rules shaping a wide range of human behavior. Understanding these rules and codifying them in computational models can provide important missing insights into why systems function the way they do (static) as well as the pace and structure of transitions required to improve sustainability (dynamic). Here, we conduct a comparative synthesis of three modeling approaches- integrated assessment modeling, engineering-economic optimization, and agent-based modeling-with underexplored potential to represent institutions. We first perform modeling experiments on climate mitigation systems that represent specific aspects of heterogeneous institutions, including formal policies and institutional coordination, and informal attitudes and norms. We find measurable but uneven aggregate impacts, while more politically meaningful distributional impacts are large across various actors. Our results show that omitting institutions can influence the costs of climate mitigation and miss opportunities to leverage institutional forces to speed up emissions reduction. These experiments allow us to explore the capacity of each modeling approach to represent insitutions and to lay out a vision for the next frontier of endogenizing institutional change in sustainability science models. To bridge the gap between modeling, theories, and empirical evidence on social institutions, this research agenda calls for joint efforts between sustainability modelers who wish to explore and incorporate institutional detail, and social scientists studying the socio-political and economic foundations for sustainability transitions.

A synthetic agent ameliorates polycystic kidney disease by promoting apoptosis of cystic cells through increased oxidative stress.

Autosomal dominant polycystic kidney disease (ADPKD) is the most common monogenic cause of chronic kidney disease and the fourth leading cause of end-stage kidney disease, accounting for over 50% of prevalent cases requiring renal replacement therapy. There is a pressing need for improved therapy for ADPKD. Recent insights into the pathophysiology of ADPKD revealed that cyst cells undergo metabolic changes that up-regulate aerobic glycolysis in lieu of mitochondrial respiration for energy production, a process that ostensibly fuels their increased proliferation. The present work leverages this metabolic disruption as a way to selectively target cyst cells for apoptosis. This small-molecule therapeutic strategy utilizes 11beta-dichloro, a repurposed DNA-damaging anti-tumor agent that induces apoptosis by exacerbating mitochondrial oxidative stress. Here, we demonstrate that 11beta-dichloro is effective in delaying cyst growth and its associated inflammatory and fibrotic events, thus preserving kidney function in perinatal and adult mouse models of ADPKD. In both models, the cyst cells with homozygous inactivation of Pkd1 show enhanced oxidative stress following treatment with 11beta-dichloro and undergo apoptosis. Co-administration of the antioxidant vitamin E negated the therapeutic benefit of 11beta-dichloro in vivo, supporting the conclusion that oxidative stress is a key component of the mechanism of action. As a preclinical development primer, we also synthesized and tested an 11beta-dichloro derivative that cannot directly alkylate DNA, while retaining pro-oxidant features. This derivative nonetheless maintains excellent anti-cystic properties in vivo and emerges as the lead candidate for development.

Antibody design using deep learning: from sequence and structure design to affinity maturation.

Deep learning has achieved impressive results in various fields such as computer vision and natural language processing, making it a powerful tool in biology. Its applications now encompass cellular image classification, genomic studies and drug discovery. While drug development traditionally focused deep learning applications on small molecules, recent innovations have incorporated it in the discovery and development of biological molecules, particularly antibodies. Researchers have devised novel techniques to streamline antibody development, combining in vitro and in silico methods. In particular, computational power expedites lead candidate generation, scaling and potential antibody development against complex antigens. This survey highlights significant advancements in protein design and optimization, specifically focusing on antibodies. This includes various aspects such as design, folding, antibody-antigen interactions docking and affinity maturation.

ChemMORT: an automatic ADMET optimization platform using deep learning and multi-objective particle swarm optimization.

Drug discovery and development constitute a laborious and costly undertaking. The success of a drug hinges not only good efficacy but also acceptable absorption, distribution, metabolism, elimination, and toxicity (ADMET) properties. Overall, up to 50% of drug development failures have been contributed from undesirable ADMET profiles. As a multiple parameter objective, the optimization of the ADMET properties is extremely challenging owing to the vast chemical space and limited human expert knowledge. In this study, a freely available platform called Chemical Molecular Optimization, Representation and Translation (ChemMORT) is developed for the optimization of multiple ADMET endpoints without the loss of potency (https://cadd.nscc-tj.cn/deploy/chemmort/). ChemMORT contains three modules: Simplified Molecular Input Line Entry System (SMILES) Encoder, Descriptor Decoder and Molecular Optimizer. The SMILES Encoder can generate the molecular representation with a 512-dimensional vector, and the Descriptor Decoder is able to translate the above representation to the corresponding molecular structure with high accuracy. Based on reversible molecular representation and particle swarm optimization strategy, the Molecular Optimizer can be used to effectively optimize undesirable ADMET properties without the loss of bioactivity, which essentially accomplishes the design of inverse QSAR. The constrained multi-objective optimization of the poly (ADP-ribose) polymerase-1 inhibitor is provided as the case to explore the utility of ChemMORT.

AttABseq: an attention-based deep learning prediction method for antigen-antibody binding affinity changes based on protein sequences.

The optimization of therapeutic antibodies through traditional techniques, such as candidate screening via hybridoma or phage display, is resource-intensive and time-consuming. In recent years, computational and artificial intelligence-based methods have been actively developed to accelerate and improve the development of therapeutic antibodies. In this study, we developed an end-to-end sequence-based deep learning model, termed AttABseq, for the predictions of the antigen-antibody binding affinity changes connected with antibody mutations. AttABseq is a highly efficient and generic attention-based model by utilizing diverse antigen-antibody complex sequences as the input to predict the binding affinity changes of residue mutations. The assessment on the three benchmark datasets illustrates that AttABseq is 120% more accurate than other sequence-based models in terms of the Pearson correlation coefficient between the predicted and experimental binding affinity changes. Moreover, AttABseq also either outperforms or competes favorably with the structure-based approaches. Furthermore, AttABseq consistently demonstrates robust predictive capabilities across a diverse array of conditions, underscoring its remarkable capacity for generalization across a wide spectrum of antigen-antibody complexes. It imposes no constraints on the quantity of altered residues, rendering it particularly applicable in scenarios where crystallographic structures remain unavailable. The attention-based interpretability analysis indicates that the causal effects of point mutations on antibody-antigen binding affinity changes can be visualized at the residue level, which might assist automated antibody sequence optimization. We believe that AttABseq provides a fiercely competitive answer to therapeutic antibody optimization.

A kinetic model for solving a combination optimization problem in ab-initio Cryo-EM 3D reconstruction.

Cryo-Electron Microscopy (cryo-EM) is a widely used and effective method for determining the three-dimensional (3D) structure of biological molecules. For ab-initio Cryo-EM 3D reconstruction using single particle analysis (SPA), estimating the projection direction of the projection image is a crucial step. However, the existing SPA methods based on common lines are sensitive to noise. The error in common line detection will lead to a poor estimation of the projection directions and thus may greatly affect the final reconstruction results. To improve the reconstruction results, multiple candidate common lines are estimated for each pair of projection images. The key problem then becomes a combination optimization problem of selecting consistent common lines from multiple candidates. To solve the problem efficiently, a physics-inspired method based on a kinetic model is proposed in this work. More specifically, hypothetical attractive forces between each pair of candidate common lines are used to calculate a hypothetical torque exerted on each projection image in the 3D reconstruction space, and the rotation under the hypothetical torque is used to optimize the projection direction estimation of the projection image. This way, the consistent common lines along with the projection directions can be found directly without enumeration of all the combinations of the multiple candidate common lines. Compared with the traditional methods, the proposed method is shown to be able to produce more accurate 3D reconstruction results from high noise projection images. Besides the practical value, the proposed method also serves as a good reference for solving similar combinatorial optimization problems.

MM optimization: Proximal distance algorithms, path following, and trust regions.

We briefly review the majorization-minimization (MM) principle and elaborate on the closely related notion of proximal distance algorithms, a generic approach for solving constrained optimization problems via quadratic penalties. We illustrate how the MM and proximal distance principles apply to a variety of problems from statistics, finance, and nonlinear optimization. Drawing from our selected examples, we also sketch a few ideas pertinent to the acceleration of MM algorithms: a) structuring updates around efficient matrix decompositions, b) path following in proximal distance iteration, and c) cubic majorization and its connections to trust region methods. These ideas are put to the test on several numerical examples, but for the sake of brevity, we omit detailed comparisons to competing methods. The current article, which is a mix of review and current contributions, celebrates the MM principle as a powerful framework for designing optimization algorithms and reinterpreting existing ones.

Barriers for the performance of graph neural networks (GNN) in discrete random structures.

Recently, graph neural network (GNN)-based algorithms were proposed to solve a variety of combinatorial optimization problems [M. J. Schuetz, J. K. Brubaker, H. G. Katzgraber, Nat. Mach. Intell.4, 367-377 (2022)]. GNN was tested in particular on randomly generated instances of these problems. The publication [M. J. Schuetz, J. K. Brubaker, H. G. Katzgraber, Nat. Mach. Intell.4, 367-377 (2022)] stirred a debate whether the GNN-based method was adequately benchmarked against best prior methods. In particular, critical commentaries [M. C. Angelini, F. Ricci-Tersenghi, Nat. Mach. Intell.5, 29-31 (2023)] and [S. Boettcher, Nat. Mach. Intell.5, 24-25 (2023)] point out that a simple greedy algorithm performs better than the GNN. We do not intend to discuss the merits of arguments and counterarguments in these papers. Rather, in this note, we establish a fundamental limitation for running GNN on random instances considered in these references, for a broad range of choices of GNN architecture. Specifically, these barriers hold when the depth of GNN does not scale with graph size (we note that depth 2 was used in experiments in [M. J. Schuetz, J. K. Brubaker, H. G. Katzgraber, Nat. Mach. Intell.4, 367-377 (2022)]), and importantly, these barriers hold regardless of any other parameters of GNN architecture. These limitations arise from the presence of the overlap gap property (OGP) phase transition, which is a barrier for many algorithms, including importantly local algorithms, of which GNN is an example. At the same time, some algorithms known prior to the introduction of GNN provide best results for these problems up to the OGP phase transition. This leaves very little space for GNN to outperform the known algorithms, and based on this, we side with the conclusions made in [M. C. Angelini, F. Ricci-Tersenghi, Nat. Mach. Intell.5, 29-31 (2023)] and [S. Boettcher, Nat. Mach. Intell.5, 24-25 (2023)].

Humans plan for the near future to walk economically on uneven terrain.

Humans experience small fluctuations in their gait when walking on uneven terrain. The fluctuations deviate from the steady, energy-minimizing pattern for level walking and have no obvious organization. But humans often look ahead when they walk, and could potentially plan anticipatory fluctuations for the terrain. Such planning is only sensible if it serves some an objective purpose, such as maintaining constant speed or reducing energy expenditure, that is also attainable within finite planning capacity. Here, we show that humans do plan and perform optimal control strategies on uneven terrain. Rather than maintaining constant speed, they make purposeful, anticipatory speed adjustments that are consistent with minimizing energy expenditure. A simple optimal control model predicts economical speed fluctuations that agree well with experiments with humans (N = 12) walking on seven different terrain profiles (correlated with model [Formula: see text] , [Formula: see text] all terrains). Participants made repeatable speed fluctuations starting about six to eight steps ahead of each terrain feature (up to ±7.5 cm height difference each step, up to 16 consecutive features). Nearer features matter more, because energy is dissipated with each succeeding step's collision with ground, preventing momentum from persisting indefinitely. A finite horizon of continuous look-ahead and motor working space thus suffice to practically optimize for any length of terrain. Humans reason about walking in the near future to plan complex optimal control sequences.

Behavioral signatures of face perception emerge in deep neural networks optimized for face recognition.

Human face recognition is highly accurate and exhibits a number of distinctive and well-documented behavioral "signatures" such as the use of a characteristic representational space, the disproportionate performance cost when stimuli are presented upside down, and the drop in accuracy for faces from races the participant is less familiar with. These and other phenomena have long been taken as evidence that face recognition is "special". But why does human face perception exhibit these properties in the first place? Here, we use deep convolutional neural networks (CNNs) to test the hypothesis that all of these signatures of human face perception result from optimization for the task of face recognition. Indeed, as predicted by this hypothesis, these phenomena are all found in CNNs trained on face recognition, but not in CNNs trained on object recognition, even when additionally trained to detect faces while matching the amount of face experience. To test whether these signatures are in principle specific to faces, we optimized a CNN on car discrimination and tested it on upright and inverted car images. As we found for face perception, the car-trained network showed a drop in performance for inverted vs. upright cars. Similarly, CNNs trained on inverted faces produced an inverted face inversion effect. These findings show that the behavioral signatures of human face perception reflect and are well explained as the result of optimization for the task of face recognition, and that the nature of the computations underlying this task may not be so special after all.

James-Stein for the leading eigenvector.

Recent research identifies and corrects bias, such as excess dispersion, in the leading sample eigenvector of a factor-based covariance matrix estimated from a high-dimension low sample size (HL) data set. We show that eigenvector bias can have a substantial impact on variance-minimizing optimization in the HL regime, while bias in estimated eigenvalues may have little effect. We describe a data-driven eigenvector shrinkage estimator in the HL regime called "James-Stein for eigenvectors" (JSE) and its close relationship with the James-Stein (JS) estimator for a collection of averages. We show, both theoretically and with numerical experiments, that, for certain variance-minimizing problems of practical importance, efforts to correct eigenvalues have little value in comparison to the JSE correction of the leading eigenvector. When certain extra information is present, JSE is a consistent estimator of the leading eigenvector.

Benefits and trade-offs of optimizing global land use for food, water, and carbon.

Current large-scale patterns of land use reflect history, local traditions, and production costs, much more so than they reflect biophysical potential or global supply and demand for food and freshwater, or-more recently-climate change mitigation. We quantified alternative land-use allocations that consider trade-offs for these demands by combining a dynamic vegetation model and an optimization algorithm to determine Pareto-optimal land-use allocations under changing climate conditions in 2090-2099 and alternatively in 2033-2042. These form the outer bounds of the option space for global land-use transformation. Results show a potential to increase all three indicators (+83% in crop production, +8% in available runoff, and +3% in carbon storage globally) compared to the current land-use configuration, with clear land-use priority areas: Tropical and boreal forests were preserved, crops were produced in temperate regions, and pastures were preferentially allocated in semiarid grasslands and savannas. Transformations toward optimal land-use patterns would imply extensive reconfigurations and changes in land management, but the required annual land-use changes were nevertheless of similar magnitude as those suggested by established land-use change scenarios. The optimization results clearly show that large benefits could be achieved when land use is reconsidered under a "global supply" perspective with a regional focus that differs across the world's regions in order to achieve the supply of key ecosystem services under the emerging global pressures.