Prediction of whole-cell transcriptional response with machine learning.

MOTIVATION: Applications in synthetic and systems biology can benefit from measuring whole-cell response to biochemical perturbations. Execution of experiments to cover all possible combinations of perturbations is infeasible. In this paper, we present the host response model (HRM), a machine learning approach that maps response of single perturbations to transcriptional response of the combination of perturbations. RESULTS: The HRM combines high-throughput sequencing with machine learning to infer links between experimental context, prior knowledge of cell regulatory networks, and RNASeq data to predict a gene's dysregulation. We find that the HRM can predict the directionality of dysregulation to a combination of inducers with an accuracy of >90% using data from single inducers. We further find that the use of prior, known cell regulatory networks doubles the predictive performance of the HRM (an R2 from 0.3 to 0.65). The model was validated in two organisms, Escherichia coli and Bacillus subtilis, using new experiments conducted after training. Finally, while the HRM is trained with gene expression data, the direct prediction of differential expression makes it possible to also conduct enrichment analyses using its predictions. We show that the HRM can accurately classify >95% of the pathway regulations. The HRM reduces the number of RNASeq experiments needed as responses can be tested in silico prior to the experiment. AVAILABILITY AND IMPLEMENTATION: The HRM software and tutorial are available at https://github.com/sd2e/CDM and the configurable differential expression analysis tools and tutorials are available at https://github.com/SD2E/omics_tools. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Biosynthesis of the nitrogenase active-site cofactor precursor NifB-co in Saccharomyces cerevisiae.

The radical S-adenosylmethionine (SAM) enzyme NifB occupies a central and essential position in nitrogenase biogenesis. NifB catalyzes the formation of an [8Fe-9S-C] cluster, called NifB-co, which constitutes the core of the active-site cofactors for all 3 nitrogenase types. Here, we produce functional NifB in aerobically cultured Saccharomyces cerevisiae Combinatorial pathway design was employed to construct 62 strains in which transcription units driving different expression levels of mitochondria-targeted nif genes (nifUSXB and fdxN) were integrated into the chromosome. Two combinatorial libraries totaling 0.7 Mb were constructed: An expression library of 6 partial clusters, including nifUSX and fdxN, and a library consisting of 28 different nifB genes mined from the Structure-Function Linkage Database and expressed at different levels according to a factorial design. We show that coexpression in yeast of the nitrogenase maturation proteins NifU, NifS, and FdxN from Azotobacter vinelandii with NifB from the archaea Methanocaldococcus infernus or Methanothermobacter thermautotrophicus yields NifB proteins equipped with [Fe-S] clusters that, as purified, support in vitro formation of NifB-co. Proof of in vivo NifB-co formation was additionally obtained. NifX as purified from aerobically cultured S. cerevisiae coexpressing M. thermautotrophicus NifB with A. vinelandii NifU, NifS, and FdxN, and engineered yeast SAM synthase supported FeMo-co synthesis, indicative of NifX carrying in vivo-formed NifB-co. This study defines the minimal genetic determinants for the formation of the key precursor in the nitrogenase cofactor biosynthetic pathway in a eukaryotic organism.

Registry in a tube: multiplexed pools of retrievable parts for genetic design space exploration.

Genetic designs can consist of dozens of genes and hundreds of genetic parts. After evaluating a design, it is desirable to implement changes without the cost and burden of starting the construction process from scratch. Here, we report a two-step process where a large design space is divided into deep pools of composite parts, from which individuals are retrieved and assembled to build a final construct. The pools are built via multiplexed assembly and sequenced using next-generation sequencing. Each pool consists of ∼20 Mb of up to 5000 unique and sequence-verified composite parts that are barcoded for retrieval by PCR. This approach is applied to a 16-gene nitrogen fixation pathway, which is broken into pools containing a total of 55 848 composite parts (71.0 Mb). The pools encompass an enormous design space (1043 possible 23 kb constructs), from which an algorithm-guided 192-member 4.5 Mb library is built. Next, all 1030 possible genetic circuits based on 10 repressors (NOR/NOT gates) are encoded in pools where each repressor is fused to all permutations of input promoters. These demonstrate that multiplexing can be applied to encompass entire design spaces from which individuals can be accessed and evaluated.

A Pressure Test to Make 10 Molecules in 90 Days: External Evaluation of Methods to Engineer Biology.

Centralized facilities for genetic engineering, or "biofoundries", offer the potential to design organisms to address emerging needs in medicine, agriculture, industry, and defense. The field has seen rapid advances in technology, but it is difficult to gauge current capabilities or identify gaps across projects. To this end, our foundry was assessed via a timed "pressure test", in which 3 months were given to build organisms to produce 10 molecules unknown to us in advance. By applying a diversity of new approaches, we produced the desired molecule or a closely related one for six out of 10 targets during the performance period and made advances toward production of the others as well. Specifically, we increased the titers of 1-hexadecanol, pyrrolnitrin, and pacidamycin D, found novel routes to the enediyne warhead underlying powerful antimicrobials, established a cell-free system for monoterpene production, produced an intermediate toward vincristine biosynthesis, and encoded 7802 individually retrievable pathways to 540 bisindoles in a DNA pool. Pathways to tetrahydrofuran and barbamide were designed and constructed, but toxicity or analytical tools inhibited further progress. In sum, we constructed 1.2 Mb DNA, built 215 strains spanning five species ( Saccharomyces cerevisiae, Escherichia coli, Streptomyces albidoflavus, Streptomyces coelicolor, and Streptomyces albovinaceus), established two cell-free systems, and performed 690 assays developed in-house for the molecules.

Genetic circuit characterization and debugging using RNA-seq.

Genetic circuits implement computational operations within a cell. Debugging them is difficult because their function is defined by multiple states (e.g., combinations of inputs) that vary in time. Here, we develop RNA-seq methods that enable the simultaneous measurement of: (i) the states of internal gates, (ii) part performance (promoters, insulators, terminators), and (iii) impact on host gene expression. This is applied to a three-input one-output circuit consisting of three sensors, five NOR/NOT gates, and 46 genetic parts. Transcription profiles are obtained for all eight combinations of inputs, from which biophysical models can extract part activities and the response functions of sensors and gates. Various unexpected failure modes are identified, including cryptic antisense promoters, terminator failure, and a sensor malfunction due to media-induced changes in host gene expression. This can guide the selection of new parts to fix these problems, which we demonstrate by using a bidirectional terminator to disrupt observed antisense transcription. This work introduces RNA-seq as a powerful method for circuit characterization and debugging that overcomes the limitations of fluorescent reporters and scales to large systems composed of many parts.

Histone H4K20/H3K9 demethylase PHF8 regulates zebrafish brain and craniofacial development.

X-linked mental retardation (XLMR) is a complex human disease that causes intellectual disability. Causal mutations have been found in approximately 90 X-linked genes; however, molecular and biological functions of many of these genetically defined XLMR genes remain unknown. PHF8 (PHD (plant homeo domain) finger protein 8) is a JmjC domain-containing protein and its mutations have been found in patients with XLMR and craniofacial deformities. Here we provide multiple lines of evidence establishing PHF8 as the first mono-methyl histone H4 lysine 20 (H4K20me1) demethylase, with additional activities towards histone H3K9me1 and me2. PHF8 is located around the transcription start sites (TSS) of approximately 7,000 RefSeq genes and in gene bodies and intergenic regions (non-TSS). PHF8 depletion resulted in upregulation of H4K20me1 and H3K9me1 at the TSS and H3K9me2 in the non-TSS sites, respectively, demonstrating differential substrate specificities at different target locations. PHF8 positively regulates gene expression, which is dependent on its H3K4me3-binding PHD and catalytic domains. Importantly, patient mutations significantly compromised PHF8 catalytic function. PHF8 regulates cell survival in the zebrafish brain and jaw development, thus providing a potentially relevant biological context for understanding the clinical symptoms associated with PHF8 patients. Lastly, genetic and molecular evidence supports a model whereby PHF8 regulates zebrafish neuronal cell survival and jaw development in part by directly regulating the expression of the homeodomain transcription factor MSX1/MSXB, which functions downstream of multiple signalling and developmental pathways. Our findings indicate that an imbalance of histone methylation dynamics has a critical role in XLMR.

Optimized detection of sequence variation in heterozygous genomes using DNA microarrays with isothermal-melting probes.

The use of DNA microarrays to identify nucleotide variation is almost 20 years old. A variety of improvements in probe design and experimental conditions have brought this technology to the point that single-nucleotide differences can be efficiently detected in unmixed samples, although developing reliable methods for detection of mixed sequences (e.g., heterozygotes) remains challenging. Surprisingly, a comprehensive study of the probe design parameters and experimental conditions that optimize discrimination of single-nucleotide polymorphisms (SNPs) has yet to be reported, so the limits of this technology remain uncertain. By targeting 24,549 SNPs that differ between two Saccharomyces cerevisiae strains, we studied the effect of SNPs on hybridization efficiency to DNA microarray probes of different lengths under different hybridization conditions. We found that the critical parameter for optimization of sequence discrimination is the relationship between probe melting temperature (T(m)) and the temperature at which the hybridization reaction is performed. This relationship can be exploited through the design of microarrays containing probes of equal T(m) by varying the length of probes. We demonstrate using such a microarray that we detect >90% homozygous SNPs and >80% heterozygous SNPs using the SNPScanner algorithm. The optimized design and experimental parameters determined in this study should guide DNA microarray designs for applications that require sequence discrimination such as mutation detection, genotyping of unmixed and mixed samples, and allele-specific gene expression. Moreover, designing microarray probes with optimized sensitivity to mismatches should increase the accuracy of standard microarray applications such as copy-number variation detection and gene expression analysis.

Parallel engineering of environmental bacteria and performance over years under jungle-simulated conditions.

Engineered bacteria could perform many functions in the environment, for example, to remediate pollutants, deliver nutrients to crops or act as in-field biosensors. Model organisms can be unreliable in the field, but selecting an isolate from the thousands that naturally live there and genetically manipulating them to carry the desired function is a slow and uninformed process. Here, we demonstrate the parallel engineering of isolates from environmental samples by using the broad-host-range XPORT conjugation system (Bacillus subtilis mini-ICEBs1) to transfer a genetic payload to many isolates in parallel. Bacillus and Lysinibacillus species were obtained from seven soil and water samples from different locations in Israel. XPORT successfully transferred a genetic function (reporter expression) into 25 of these isolates. They were then screened to identify the best-performing chassis based on the expression level, doubling time, functional stability in soil, and environmentally-relevant traits of its closest annotated reference species, such as the ability to sporulate and temperature tolerance. From this library, we selected Bacillus frigoritolerans A3E1, re-introduced it to soil, and measured function and genetic stability in a contained environment that replicates jungle conditions. After 21 months of storage, the engineered bacteria were viable, could perform their function, and did not accumulate disruptive mutations.

A toolkit for enhanced reproducibility of RNASeq analysis for synthetic biologists.

Sequencing technologies, in particular RNASeq, have become critical tools in the design, build, test and learn cycle of synthetic biology. They provide a better understanding of synthetic designs, and they help identify ways to improve and select designs. While these data are beneficial to design, their collection and analysis is a complex, multistep process that has implications on both discovery and reproducibility of experiments. Additionally, tool parameters, experimental metadata, normalization of data and standardization of file formats present challenges that are computationally intensive. This calls for high-throughput pipelines expressly designed to handle the combinatorial and longitudinal nature of synthetic biology. In this paper, we present a pipeline to maximize the analytical reproducibility of RNASeq for synthetic biologists. We also explore the impact of reproducibility on the validation of machine learning models. We present the design of a pipeline that combines traditional RNASeq data processing tools with structured metadata tracking to allow for the exploration of the combinatorial design in a high-throughput and reproducible manner. We then demonstrate utility via two different experiments: a control comparison experiment and a machine learning model experiment. The first experiment compares datasets collected from identical biological controls across multiple days for two different organisms. It shows that a reproducible experimental protocol for one organism does not guarantee reproducibility in another. The second experiment quantifies the differences in experimental runs from multiple perspectives. It shows that the lack of reproducibility from these different perspectives can place an upper bound on the validation of machine learning models trained on RNASeq data. Graphical Abstract.

Four-Step Pathway from Phenylpyruvate to Benzylamine, an Intermediate to the High-Energy Propellant CL-20.

Benzylamine is a commodity chemical used in the synthesis of motion-sickness treatments and anticonvulsants, in dyeing textiles, and as a precursor to the high-energy propellant CL-20. Because chemical production generates toxic waste streams, biosynthetic alternatives have been explored, recently resulting in a functional nine-step pathway from central metabolism (phenylalanine) in E. coli. We report a novel four-step pathway for benzylamine production, which generates the product from cellular phenylpyruvate using enzymes from different sources: a mandelate synthase (Amycolatopsis orientalis), a mandelate oxidase (Streptomyces coelicolor), a benzoylformate decarboxylase (Pseudomonas putida), and an aminotransferase (Salicibacter pomeroyi). This pathway produces benzylamine at 24 mg/L in 15 h (4.5% yield) in cultures of unoptimized cells supplemented with phenylpyruvate. Because the yield is low, supplementation with pathway intermediates is used to troubleshoot the design. This identifies conversion inefficiencies in the mandelate synthase-mediated synthesis of (S)-mandelic acid, and subsequent genome mining identifies a new mandelate synthase (Streptomyces sp. 1114.5) with improved yield. Supplementation experiments also reveal native redirection of ambient phenylpyruvate away from the pathway to phenylalanine. Overall, this work illustrates how retrosynthetic design can dramatically reduce the number of enzymes in a pathway, potentially reducing its draw on cellular resources. However, it also shows that such benefits can be abrogated by inefficiencies of individual conversions. Addressing these barriers can provide an alternative approach to green production of benzylamine, eliminating upstream dependence on chlorination chemistry.

Selection for constrained peptides that bind to a single target protein.

Peptide secondary metabolites are common in nature and have diverse pharmacologically-relevant functions, from antibiotics to cross-kingdom signaling. Here, we present a method to design large libraries of modified peptides in Escherichia coli and screen them in vivo to identify those that bind to a single target-of-interest. Constrained peptide scaffolds were produced using modified enzymes gleaned from microbial RiPP (ribosomally synthesized and post-translationally modified peptide) pathways and diversified to build large libraries. The binding of a RiPP to a protein target leads to the intein-catalyzed release of an RNA polymerase σ factor, which drives the expression of selectable markers. As a proof-of-concept, a selection was performed for binding to the SARS-CoV-2 Spike receptor binding domain. A 1625 Da constrained peptide (AMK-1057) was found that binds with similar affinity (990 ± 5 nM) as an ACE2-derived peptide. This demonstrates a generalizable method to identify constrained peptides that adhere to a single protein target, as a step towards "molecular glues" for therapeutics and diagnostics.

Transcriptional regulatory code of a eukaryotic genome.

DNA-binding transcriptional regulators interpret the genome's regulatory code by binding to specific sequences to induce or repress gene expression. Comparative genomics has recently been used to identify potential cis-regulatory sequences within the yeast genome on the basis of phylogenetic conservation, but this information alone does not reveal if or when transcriptional regulators occupy these binding sites. We have constructed an initial map of yeast's transcriptional regulatory code by identifying the sequence elements that are bound by regulators under various conditions and that are conserved among Saccharomyces species. The organization of regulatory elements in promoters and the environment-dependent use of these elements by regulators are discussed. We find that environment-specific use of regulatory elements predicts mechanistic models for the function of a large population of yeast's transcriptional regulators.

Organism Engineering for the Bioproduction of the Triaminotrinitrobenzene (TATB) Precursor Phloroglucinol (PG).

Organism engineering requires the selection of an appropriate chassis, editing its genome, combining traits from different source species, and controlling genes with synthetic circuits. When a strain is needed for a new target objective, for example, to produce a chemical-of-need, the best strains, genes, techniques, software, and expertise may be distributed across laboratories. Here, we report a project where we were assigned phloroglucinol (PG) as a target, and then combined unique capabilities across the United States Army, Navy, and Air Force service laboratories with the shared goal of designing an organism to produce this molecule. In addition to the laboratory strain Escherichia coli, organisms were screened from soil and seawater. Putative PG-producing enzymes were mined from a strain bank of bacteria isolated from aircraft and fuel depots. The best enzyme was introduced into the ocean strain Marinobacter atlanticus CP1 with its genome edited to redirect carbon flux from natural fatty acid ester (FAE) production. PG production was also attempted in Bacillus subtilis and Clostridium acetobutylicum. A genetic circuit was constructed in E. coli that responds to PG accumulation, which was then ported to an in vitro paper-based system that could serve as a platform for future low-cost strain screening or for in-field sensing. Collectively, these efforts show how distributed biotechnology laboratories with domain-specific expertise can be marshalled to quickly provide a solution for a targeted organism engineering project, and highlights data and material sharing protocols needed to accelerate future efforts.

Computational discovery of gene modules and regulatory networks.

We describe an algorithm for discovering regulatory networks of gene modules, GRAM (Genetic Regulatory Modules), that combines information from genome-wide location and expression data sets. A gene module is defined as a set of coexpressed genes to which the same set of transcription factors binds. Unlike previous approaches that relied primarily on functional information from expression data, the GRAM algorithm explicitly links genes to the factors that regulate them by incorporating DNA binding data, which provide direct physical evidence of regulatory interactions. We use the GRAM algorithm to describe a genome-wide regulatory network in Saccharomyces cerevisiae using binding information for 106 transcription factors profiled in rich medium conditions data from over 500 expression experiments. We also present a genome-wide location analysis data set for regulators in yeast cells treated with rapamycin, and use the GRAM algorithm to provide biological insights into this regulatory network

Control of type III protein secretion using a minimal genetic system.

Gram-negative bacteria secrete proteins using a type III secretion system (T3SS), which functions as a needle-like molecular machine. The many proteins involved in T3SS construction are tightly regulated due to its role in pathogenesis and motility. Here, starting with the 35 kb Salmonella pathogenicity island 1 (SPI-1), we eliminated internal regulation and simplified the genetics by removing or recoding genes, scrambling gene order and replacing all non-coding DNA with synthetic genetic parts. This process results in a 16 kb cluster that shares no sequence identity, regulation or organizational principles with SPI-1. Building this simplified system led to the discovery of essential roles for an internal start site (SpaO) and small RNA (InvR). Further, it can be controlled using synthetic regulatory circuits, including under SPI-1 repressing conditions. This work reveals an incredible post-transcriptional robustness in T3SS assembly and aids its control as a tool in biotechnology.

DNAplotlib: Programmable Visualization of Genetic Designs and Associated Data.

DNAplotlib ( www.dnaplotlib.org ) is a computational toolkit for the programmable visualization of highly customizable, standards-compliant genetic designs. Functions are provided to aid with both visualization tasks and to extract and overlay associated experimental data. High-quality output is produced in the form of vector-based PDFs, rasterized images, and animated movies. All aspects of the rendering process can be easily customized or extended by the user to cover new forms of genetic part or regulation. DNAplotlib supports improved communication of genetic design information and offers new avenues for static, interactive and dynamic visualizations that map and explore the links between the structure and function of genetic parts, devices and systems; including metabolic pathways and genetic circuits. DNAplotlib is cross-platform software developed using Python.

An improved map of conserved regulatory sites for Saccharomyces cerevisiae.

BACKGROUND: The regulatory map of a genome consists of the binding sites for proteins that determine the transcription of nearby genes. An initial regulatory map for S. cerevisiae was recently published using six motif discovery programs to analyze genome-wide chromatin immunoprecipitation data for 203 transcription factors. The programs were used to identify sequence motifs that were likely to correspond to the DNA-binding specificity of the immunoprecipitated proteins. We report improved versions of two conservation-based motif discovery algorithms, PhyloCon and Converge. Using these programs, we create a refined regulatory map for S. cerevisiae by reanalyzing the same chromatin immunoprecipitation data. RESULTS: Applying the same conservative criteria that were applied in the original study, we find that PhyloCon and Converge each separately discover more known specificities than the combination of all six programs in the previous study. Combining the results of PhyloCon and Converge, we discover significant sequence motifs for 36 transcription factors that were previously missed. The new set of motifs identifies 636 more regulatory interactions than the previous one. The new network contains 28% more regulatory interactions among transcription factors, evidence of greater cross-talk between regulators. CONCLUSION: Combining two complementary computational strategies for conservation-based motif discovery improves the ability to identify the specificity of transcriptional regulators from genome-wide chromatin immunoprecipitation data. The increased sensitivity of these methods significantly expands the map of yeast regulatory sites without the need to alter any of the thresholds for statistical significance. The new map of regulatory sites reveals a more elaborate and complex view of the yeast genetic regulatory network than was observed previously.

Double Dutch: A Tool for Designing Combinatorial Libraries of Biological Systems.

Recently, semirational approaches that rely on combinatorial assembly of characterized DNA components have been used to engineer biosynthetic pathways. In practice, however, it is not practical to assemble and test millions of pathway variants in order to elucidate how different DNA components affect the behavior of a pathway. To address this challenge, we apply a rigorous mathematical approach known as design of experiments (DOE) that can be used to construct empirical models of system behavior without testing all variants. To support this approach, we have developed a tool named Double Dutch, which uses a formal grammar and heuristic algorithms to automate the process of DOE library design. Compared to designing by hand, Double Dutch enables users to more efficiently and scalably design libraries of pathway variants that can be used in a DOE framework and uniquely provides a means to flexibly balance design considerations of statistical analysis, construction cost, and risk of homologous recombination, thereby demonstrating the utility of automating decision making when faced with complex design trade-offs.

Memory and Combinatorial Logic Based on DNA Inversions: Dynamics and Evolutionary Stability.

Genetic memory can be implemented using enzymes that catalyze DNA inversions, where each orientation corresponds to a "bit". Here, we use two DNA invertases (FimE and HbiF) that reorient DNA irreversibly between two states with opposite directionality. First, we construct memory that is set by FimE and reset by HbiF. Next, we build a NOT gate where the input promoter drives FimE and in the absence of signal the reverse state is maintained by the constitutive expression of HbiF. The gate requires ∼3 h to turn on and off. The evolutionary stabilities of these circuits are measured by passaging cells while cycling function. The memory switch is stable over 400 h (17 days, 14 state changes); however, the gate breaks after 54 h (>2 days) due to continuous invertase expression. Genome sequencing reveals that the circuit remains intact, but the host strain evolves to reduce invertase expression. This work highlights the need to evaluate the evolutionary robustness and failure modes of circuit designs, especially as more complex multigate circuits are implemented.