A field guide to cultivating computational biology.

Evolving in sync with the computation revolution over the past 30 years, computational biology has emerged as a mature scientific field. While the field has made major contributions toward improving scientific knowledge and human health, individual computational biology practitioners at various institutions often languish in career development. As optimistic biologists passionate about the future of our field, we propose solutions for both eager and reluctant individual scientists, institutions, publishers, funding agencies, and educators to fully embrace computational biology. We believe that in order to pave the way for the next generation of discoveries, we need to improve recognition for computational biologists and better align pathways of career success with pathways of scientific progress. With 10 outlined steps, we call on all adjacent fields to move away from the traditional individual, single-discipline investigator research model and embrace multidisciplinary, data-driven, team science.

Inferring gene regulatory networks using transcriptional profiles as dynamical attractors.

Genetic regulatory networks (GRNs) regulate the flow of genetic information from the genome to expressed messenger RNAs (mRNAs) and thus are critical to controlling the phenotypic characteristics of cells. Numerous methods exist for profiling mRNA transcript levels and identifying protein-DNA binding interactions at the genome-wide scale. These enable researchers to determine the structure and output of transcriptional regulatory networks, but uncovering the complete structure and regulatory logic of GRNs remains a challenge. The field of GRN inference aims to meet this challenge using computational modeling to derive the structure and logic of GRNs from experimental data and to encode this knowledge in Boolean networks, Bayesian networks, ordinary differential equation (ODE) models, or other modeling frameworks. However, most existing models do not incorporate dynamic transcriptional data since it has historically been less widely available in comparison to "static" transcriptional data. We report the development of an evolutionary algorithm-based ODE modeling approach (named EA) that integrates kinetic transcription data and the theory of attractor matching to infer GRN architecture and regulatory logic. Our method outperformed six leading GRN inference methods, none of which incorporate kinetic transcriptional data, in predicting regulatory connections among TFs when applied to a small-scale engineered synthetic GRN in Saccharomyces cerevisiae. Moreover, we demonstrate the potential of our method to predict unknown transcriptional profiles that would be produced upon genetic perturbation of the GRN governing a two-state cellular phenotypic switch in Candida albicans. We established an iterative refinement strategy to facilitate candidate selection for experimentation; the experimental results in turn provide validation or improvement for the model. In this way, our GRN inference approach can expedite the development of a sophisticated mathematical model that can accurately describe the structure and dynamics of the in vivo GRN.

ACTIVA: realistic single-cell RNA-seq generation with automatic cell-type identification using introspective variational autoencoders.

MOTIVATION: Single-cell RNA sequencing (scRNAseq) technologies allow for measurements of gene expression at a single-cell resolution. This provides researchers with a tremendous advantage for detecting heterogeneity, delineating cellular maps or identifying rare subpopulations. However, a critical complication remains: the low number of single-cell observations due to limitations by rarity of subpopulation, tissue degradation or cost. This absence of sufficient data may cause inaccuracy or irreproducibility of downstream analysis. In this work, we present Automated Cell-Type-informed Introspective Variational Autoencoder (ACTIVA): a novel framework for generating realistic synthetic data using a single-stream adversarial variational autoencoder conditioned with cell-type information. Within a single framework, ACTIVA can enlarge existing datasets and generate specific subpopulations on demand, as opposed to two separate models [such as single-cell GAN (scGAN) and conditional scGAN (cscGAN)]. Data generation and augmentation with ACTIVA can enhance scRNAseq pipelines and analysis, such as benchmarking new algorithms, studying the accuracy of classifiers and detecting marker genes. ACTIVA will facilitate analysis of smaller datasets, potentially reducing the number of patients and animals necessary in initial studies. RESULTS: We train and evaluate models on multiple public scRNAseq datasets. In comparison to GAN-based models (scGAN and cscGAN), we demonstrate that ACTIVA generates cells that are more realistic and harder for classifiers to identify as synthetic which also have better pair-wise correlation between genes. Data augmentation with ACTIVA significantly improves classification of rare subtypes (more than 45% improvement compared with not augmenting and 4% better than cscGAN) all while reducing run-time by an order of magnitude in comparison to both models. AVAILABILITY AND IMPLEMENTATION: The codes and datasets are hosted on Zenodo (https://doi.org/10.5281/zenodo.5879639). Tutorials are available at https://github.com/SindiLab/ACTIVA. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Modeling and Global Sensitivity Analysis of Strategies to Mitigate Covid-19 Transmission on a Structured College Campus.

In response to the COVID-19 pandemic, many higher educational institutions moved their courses on-line in hopes of slowing disease spread. The advent of multiple highly-effective vaccines offers the promise of a return to "normal" in-person operations, but it is not clear if-or for how long-campuses should employ non-pharmaceutical interventions such as requiring masks or capping the size of in-person courses. In this study, we develop and fine-tune a model of COVID-19 spread to UC Merced's student and faculty population. We perform a global sensitivity analysis to consider how both pharmaceutical and non-pharmaceutical interventions impact disease spread. Our work reveals that vaccines alone may not be sufficient to eradicate disease dynamics and that significant contact with an infectious surrounding community will maintain infections on-campus. Our work provides a foundation for higher-education planning allowing campuses to balance the benefits of in-person instruction with the ability to quarantine/isolate infectious individuals.

Computationally Driven Discovery in Coagulation.

Bleeding frequency and severity within clinical categories of hemophilia A are highly variable and the origin of this variation is unknown. Solving this mystery in coagulation requires the generation and analysis of large data sets comprised of experimental outputs or patient samples, both of which are subject to limited availability. In this review, we describe how a computationally driven approach bypasses such limitations by generating large synthetic patient data sets. These data sets were created with a mechanistic mathematical model, by varying the model inputs, clotting factor, and inhibitor concentrations, within normal physiological ranges. Specific mathematical metrics were chosen from the model output, used as a surrogate measure for bleeding severity, and statistically analyzed for further exploration and hypothesis generation. We highlight results from our recent study that employed this computationally driven approach to identify FV (factor V) as a key modifier of thrombin generation in mild to moderate hemophilia A, which was confirmed with complementary experimental assays. The mathematical model was used further to propose a potential mechanism for these observations whereby thrombin generation is rescued in FVIII-deficient plasma due to reduced substrate competition between FV and FVIII for FXa (activated factor X).

The Art and Science of Building a Computational Model to Understand Hemostasis.

Computational models of various facets of hemostasis and thrombosis have increased substantially in the last decade. These models have the potential to make predictions that can uncover new mechanisms within the complex dynamics of thrombus formation. However, these predictions are only as good as the data and assumptions they are built upon, and therefore model building requires intimate coupling with experiments. The objective of this article is to guide the reader through how a computational model is built and how it can inform and be refined by experiments. This is accomplished by answering six questions facing the model builder: (1) Why make a model? (2) What kind of model should be built? (3) How is the model built? (4) Is the model a "good" model? (5) Do we believe the model? (6) Is the model useful? These questions are answered in the context of a model of thrombus formation that has been successfully applied to understanding the interplay between blood flow, platelet deposition, and coagulation and in identifying potential modifiers of thrombin generation in hemophilia A.

A unifying model for the propagation of prion proteins in yeast brings insight into the [PSI+] prion.

The use of yeast systems to study the propagation of prions and amyloids has emerged as a crucial aspect of the global endeavor to understand those mechanisms. Yeast prion systems are intrinsically multi-scale: the molecular chemical processes are indeed coupled to the cellular processes of cell growth and division to influence phenotypical traits, observable at the scale of colonies. We introduce a novel modeling framework to tackle this difficulty using impulsive differential equations. We apply this approach to the [PSI+] yeast prion, which is associated with the misconformation and aggregation of Sup35. We build a model that reproduces and unifies previously conflicting experimental observations on [PSI+] and thus sheds light onto characteristics of the intracellular molecular processes driving aggregate replication. In particular our model uncovers a kinetic barrier for aggregate replication at low densities, meaning the change between prion or prion-free phenotype is a bi-stable transition. This result is based on the study of prion curing experiments, as well as the phenomenon of colony sectoring, a phenotype which is often ignored in experimental assays and has never been modeled. Furthermore, our results provide further insight into the effect of guanidine hydrochloride (GdnHCl) on Sup35 aggregates. To qualitatively reproduce the GdnHCl curing experiment, aggregate replication must not be completely inhibited, which suggests the existence of a mechanism different than Hsp104-mediated fragmentation. Those results are promising for further development of the [PSI+] model, but also for extending the use of this novel framework to other yeast prion or amyloid systems.

CD8 Follicular T Cells Promote B Cell Antibody Class Switch in Autoimmune Disease.

CD8 T cells can play both a protective and pathogenic role in inflammation and autoimmune development. Recent studies have highlighted the ability of CD8 T cells to function as T follicular helper (Tfh) cells in the germinal center in the context of infection. However, whether this phenomenon occurs in autoimmunity and contributes to autoimmune pathogenesis is largely unexplored. In this study, we show that CD8 T cells acquire a CD4 Tfh profile in the absence of functional regulatory T cells in both the IL-2-deficient and scurfy mouse models. Depletion of CD8 T cells mitigates autoimmune pathogenesis in IL-2-deficient mice. CD8 T cells express the B cell follicle-localizing chemokine receptor CXCR5, a principal Tfh transcription factor Bcl6, and the Tfh effector cytokine IL-21. CD8 T cells localize to the B cell follicle, express B cell costimulatory proteins, and promote B cell differentiation and Ab isotype class switching. These data reveal a novel contribution of autoreactive CD8 T cells to autoimmune disease, in part, through CD4 follicular-like differentiation and functionality.

A dominant-negative mutant inhibits multiple prion variants through a common mechanism.

Prions adopt alternative, self-replicating protein conformations and thereby determine novel phenotypes that are often irreversible. Nevertheless, dominant-negative prion mutants can revert phenotypes associated with some conformations. These observations suggest that, while intervention is possible, distinct inhibitors must be developed to overcome the conformational plasticity of prions. To understand the basis of this specificity, we determined the impact of the G58D mutant of the Sup35 prion on three of its conformational variants, which form amyloids in S. cerevisiae. G58D had been previously proposed to have unique effects on these variants, but our studies suggest a common mechanism. All variants, including those reported to be resistant, are inhibited by G58D but at distinct doses. G58D lowers the kinetic stability of the associated amyloid, enhancing its fragmentation by molecular chaperones, promoting Sup35 resolubilization, and leading to amyloid clearance particularly in daughter cells. Reducing the availability or activity of the chaperone Hsp104, even transiently, reverses curing. Thus, the specificity of inhibition is determined by the sensitivity of variants to the mutant dosage rather than mode of action, challenging the view that a unique inhibitor must be developed to combat each variant.

Assessing the impact of product inhibition in a chromogenic assay.

Chromogenic substrates (CS) are synthetic substrates used to monitor the activity of a target enzyme. It has been reported that some CSs display competitive product inhibition with their target enzyme. Thus, in assays where enzyme activity is continuously monitored over long periods of time, the product inhibition may significantly interfere with the reactions being monitored. Despite this knowledge, it is rare for CSs to be directly incorporated into mathematical models that simulate these assays. This devalues the predictive power of the models. In this study, we examined the interactions between a single enzyme, coagulation factor Xa, and its chromogenic substrate. We developed, and experimentally validated, a mathematical model of a chromogenic assay for factor Xa that explicitly included product inhibition from the CS. We employed Bayesian inference, in the form of Markov-Chain Monte Carlo, to estimate the strength of the product inhibition and other sources of uncertainty such as pipetting error and kinetic rate constants. Our model, together with carefully calibrated biochemistry experiments, allowed for full characterization of the strength and impact of product inhibition in the assay. The effect of CS product inhibition in more complex reaction mixtures was further explored using mathematical models.

Generalizing a mathematical model of prion aggregation allows strain coexistence and co-stability by including a novel misfolded species.

Prions are proteins capable of adopting misfolded conformations and transmitting these conformations to other normally folded proteins. Prions are most commonly known for causing fatal neurodegenerative diseases in mammals but are also associated with several harmless phenotypes in yeast. A distinct feature of prion propagation is the existence of different phenotypical variants, called strains. It is widely accepted that these strains correspond to different conformational states of the protein, but the mechanisms driving their interactions remain poorly understood. This study uses mathematical modeling to provide insight into this problem. We show that the classical model of prion dynamics allows at most one conformational strain to stably propagate. In order to conform to biological observations of strain coexistence and co-stability, we develop an extension of the classical model by introducing a novel prion species consistent with biological studies. Qualitative analysis of this model reveals a new variety of behavior. Indeed, it allows for stable coexistence of different strains in a wide parameter range, and it also introduces intricate initial condition dependency. These new behaviors are consistent with experimental observations of prions in both mammals and yeast. As such, our model provides a valuable tool for investigating the underlying mechanisms of prion propagation and the link between prion strains and strain specific phenotypes. The consideration of a novel prion species brings a change in perspective on prion biology and we use our model to generate hypotheses about prion infectivity.

Distinct Prion Domain Sequences Ensure Efficient Amyloid Propagation by Promoting Chaperone Binding or Processing In Vivo.

Prions are a group of proteins that can adopt a spectrum of metastable conformations in vivo. These alternative states change protein function and are self-replicating and transmissible, creating protein-based elements of inheritance and infectivity. Prion conformational flexibility is encoded in the amino acid composition and sequence of the protein, which dictate its ability not only to form an ordered aggregate known as amyloid but also to maintain and transmit this structure in vivo. But, while we can effectively predict amyloid propensity in vitro, the mechanism by which sequence elements promote prion propagation in vivo remains unclear. In yeast, propagation of the [PSI+] prion, the amyloid form of the Sup35 protein, has been linked to an oligopeptide repeat region of the protein. Here, we demonstrate that this region is composed of separable functional elements, the repeats themselves and a repeat proximal region, which are both required for efficient prion propagation. Changes in the numbers of these elements do not alter the physical properties of Sup35 amyloid, but their presence promotes amyloid fragmentation, and therefore maintenance, by molecular chaperones. Rather than acting redundantly, our observations suggest that these sequence elements make complementary contributions to prion propagation, with the repeat proximal region promoting chaperone binding to and the repeats promoting chaperone processing of Sup35 amyloid.

Culturomics as a data playground for tests of selection: Mathematical approaches to detecting selection in word use.

In biological evolution traits may rise and fall in frequency due to genetic drift, where variant frequencies change by chance, or by selection where advantageous variants will rise in frequency. The neutral model of evolution, first developed by Kimura in the 1960s, has become the standard against which selection is detected. While the balance between these two important forces - drift and selection - has been well established in biology there are other domains where the contribution of these processes is still coming together. Although the idea of natural selection has been applied to the cultural domain since the time of Darwin, it has proven more challenging to positively identify cultural traits under selection both because of a lack of established tests for selection and a lack of large cultural data sets. However, in recent years with the accumulation of large cultural data sets many cultural features from pre-history pottery to modern baby names have been shown to evolve according to the neutral theory. But there is accumulating empirical evidence from cultural processes suggesting that the neutral theory alone cannot account for all features of the data. As such, there has been a renewed interest in determining whether there is selection amidst drift. Here we analyze a subset English word frequencies, and determine whether frequency change reveals processes of selection. Inspired by the Moran and Wright-Fisher models in population genetics, we developed a neutral model of word frequency variation to assess when linguistic data appears to depart from neutral evolution. As such, our model represents a possible "test for selection" in the linguistic domain. We explore how the distribution of word use has changed for sets of words in English for more than 100 years (1901-2008) as expressed in vocabulary usage in published books, made available by Google Ngram. When comparing empirical word frequency changes to our neutral model we find pervasive and systematic departures from neutrality.

A mathematical model of the dynamics of prion aggregates with chaperone-mediated fragmentation.

Prions are proteins most commonly associated with fatal neurodegenerative diseases in mammals but are also responsible for a number of harmless heritable phenotypes in yeast. These states arise when a misfolded form of a protein appears and, rather than be removed by cellular quality control mechanisms, persists. The misfolded prion protein forms aggregates and is capable of converting normally folded protein to the misfolded state through direct interaction between the two forms. The dominant mathematical model for prion aggregate dynamics has been the nucleated polymerization model (NPM) which considers the dynamics of only the normal protein and the aggregates. However, for yeast prions the molecular chaperone Hsp104 is essential for prion propagation. Further, although mammals do not express Hsp104, experimental assays have shown Hsp104 also interacts with mammalian prion aggregates. In this study, we generalize the NPM to account for molecular chaperones and develop what we call the enzyme-limited nucleated polymerization model (ELNPM). We discuss existence, uniqueness and stability of solutions to our model and demonstrate that the NPM represents a quasi-steady-state reduction of our model. We validate the ELNPM by demonstrating agreement with experimental results on the yeast prion PSI(+) that could not be supported by the NPM. Finally, we demonstrate that, in contrast to the NPM, the ELNPM permits the coexistence of multiple prion strains.

Efficient n-gram analysis in R with cmscu.

We present a new R package, cmscu, which implements a Count-Min-Sketch with conservative updating (Cormode and Muthukrishnan Journal of Algorithms, 55(1), 58-75, 2005), and its application to n-gram analyses (Goyal et al. 2012). By writing the core implementation in C++ and exposing it to R via Rcpp, we are able to provide a memory-efficient, high-throughput, and easy-to-use library. As a proof of concept, we implemented the computationally challenging (Heafield et al. 2013) modified Kneser-Ney n-gram smoothing algorithm using cmscu as the querying engine. We then explore information density measures (Jaeger Cognitive Psychology, 61(1), 23-62, 2010) from n-gram frequencies (for n=2,3) derived from a corpus of over 2.2 million reviews provided by a Yelp, Inc. dataset. We demonstrate that these text data are at a scale beyond the reach of other more common, more general-purpose libraries available through CRAN. Using the cmscu library and the smoothing implementation, we find a positive relationship between review information density and reader review ratings. We end by highlighting the important use of new efficient tools to explore behavioral phenomena in large, relatively noisy data sets.

A Study in Nucleated Polymerization Models of Protein Aggregation.

The nucleated polymerization model is a mathematical framework that has been applied to aggregation and fragmentation processes in both the discrete and continuous setting. In particular, this model has been the canonical framework for analyzing the dynamics of protein aggregates arising in prion and amyloid diseases such as as Alzheimer's and Parkinson's disease. We present an explicit steady-state solution to the aggregate size distribution governed by the discrete nucleated polymerization equations. Steady-state solutions have been previously obtained under the assumption of continuous aggregate sizes; however, the discrete solution allows for direct computation and parameter inference, as well as facilitates estimates on the accuracy of the continuous approximation.

Modeling the distribution of enzymes on lipid vesicles: A novel framework for surface-mediated reactions in coagulation.

Blood coagulation is a network of biochemical reactions wherein dozens of proteins act collectively to initiate a rapid clotting response. Coagulation reactions are lipid-surface dependent, and this dependence is thought to help localize coagulation to the site of injury and enhance the association between reactants. Current mathematical models of coagulation either do not consider lipid as a variable or do not agree with experiments where lipid concentrations were varied. Since there is no analytic rate law that depends on lipid, only apparent rate constants can be derived from enzyme kinetic experiments. We developed a new mathematical framework for modeling enzymes reactions in the presence of lipid vesicles. Here the concentrations are such that only a fraction of the vesicles harbor bound enzymes and the rest remain empty. We call the lipid vesicles with and without enzyme TF:VIIa+ and TF:VIIa- lipid, respectively. Since substrate binds to both TF:VIIa+ and TF:VIIa- lipid, our model shows that excess empty lipid acts as a strong sink for substrate. We used our framework to derive an analytic rate equation and performed constrained optimization to estimate a single, global set of intrinsic rates for the enzyme-substrate pair. Results agree with experiments and reveal a critical lipid concentration where the conversion rate of the substrate is maximized, a phenomenon known as the template effect. Next, we included product inhibition of the enzyme and derived the corresponding rate equations, which enables kinetic studies of more complex reactions. Our combined experimental and mathematical study provides a general framework for uncovering the mechanisms by which lipid mediated reactions impact coagulation processes.

Host tracheal and intestinal microbiomes inhibit Coccidioides growth in vitro.

Coccidioidomycosis, also known as Valley fever, is a disease caused by the fungal pathogen Coccidioides. Unfortunately, patients are often misdiagnosed with bacterial pneumonia, leading to inappropriate antibiotic treatment. The soil Bacillus subtilis-like species exhibits antagonistic properties against Coccidioides in vitro; however, the antagonistic capabilities of host microbiota against Coccidioides are unexplored. We sought to examine the potential of the tracheal and intestinal microbiomes to inhibit the growth of Coccidioides in vitro. We hypothesized that an uninterrupted lawn of microbiota obtained from antibiotic-free mice would inhibit the growth of Coccidioides, while partial in vitro depletion through antibiotic disk diffusion assays would allow a niche for fungal growth. We observed that the microbiota grown on 2×GYE (GYE) and Columbia colistin and nalidixic acid with 5% sheep's blood agar inhibited the growth of Coccidioides, but microbiota grown on chocolate agar did not. Partial depletion of the microbiota through antibiotic disk diffusion revealed diminished inhibition and comparable growth of Coccidioides to controls. To characterize the bacteria grown and identify potential candidates contributing to the inhibition of Coccidioides, 16S rRNA sequencing was performed on tracheal and intestinal agar cultures and murine lung extracts. We found that the host bacteria likely responsible for this inhibition primarily included Lactobacillus and Staphylococcus. The results of this study demonstrate the potential of the host microbiota to inhibit the growth of Coccidioides in vitro and suggest that an altered microbiome through antibiotic treatment could negatively impact effective fungal clearance and allow a niche for fungal growth in vivo. IMPORTANCE: Coccidioidomycosis is caused by a fungal pathogen that invades the host lungs, causing respiratory distress. In 2019, 20,003 cases of Valley fever were reported to the CDC. However, this number likely vastly underrepresents the true number of Valley fever cases, as many go undetected due to poor testing strategies and a lack of diagnostic models. Valley fever is also often misdiagnosed as bacterial pneumonia, resulting in 60%-80% of patients being treated with antibiotics prior to an accurate diagnosis. Misdiagnosis contributes to a growing problem of antibiotic resistance and antibiotic-induced microbiome dysbiosis; the implications for disease outcomes are currently unknown. About 5%-10% of symptomatic Valley fever patients develop chronic pulmonary disease. Valley fever causes a significant financial burden and a reduced quality of life. Little is known regarding what factors contribute to the development of chronic infections and treatments for the disease are limited.

A Discrete-Time Branching Process Model of Yeast Prion Curing Curves.

The infectious agent of many neurodegenerative disorders is thought to be aggregates of prion protein, which are transmitted between cells. Recent work in yeast supports this hypothesis, but suggests that only aggregates below a critical size are transmitted efficiently. The total number of transmissible aggregates in a typical cell is a key determinant of strain infectivity. In a discrete-time branching process model of a yeast colony with prions, prion aggregates increase in size according to a Poisson process and only aggregates below a threshold size are transmitted during cell division. The total number of cells with aggregates in a growing population of yeast is expressed.

Deep learning in spatial transcriptomics: Learning from the next next-generation sequencing.

Spatial transcriptomics (ST) technologies are rapidly becoming the extension of single-cell RNA sequencing (scRNAseq), holding the potential of profiling gene expression at a single-cell resolution while maintaining cellular compositions within a tissue. Having both expression profiles and tissue organization enables researchers to better understand cellular interactions and heterogeneity, providing insight into complex biological processes that would not be possible with traditional sequencing technologies. Data generated by ST technologies are inherently noisy, high-dimensional, sparse, and multi-modal (including histological images, count matrices, etc.), thus requiring specialized computational tools for accurate and robust analysis. However, many ST studies currently utilize traditional scRNAseq tools, which are inadequate for analyzing complex ST datasets. On the other hand, many of the existing ST-specific methods are built upon traditional statistical or machine learning frameworks, which have shown to be sub-optimal in many applications due to the scale, multi-modality, and limitations of spatially resolved data (such as spatial resolution, sensitivity, and gene coverage). Given these intricacies, researchers have developed deep learning (DL)-based models to alleviate ST-specific challenges. These methods include new state-of-the-art models in alignment, spatial reconstruction, and spatial clustering, among others. However, DL models for ST analysis are nascent and remain largely underexplored. In this review, we provide an overview of existing state-of-the-art tools for analyzing spatially resolved transcriptomics while delving deeper into the DL-based approaches. We discuss the new frontiers and the open questions in this field and highlight domains in which we anticipate transformational DL applications.