18 April 2025 in Communications Biology
Although synthetic biology can produce valuable chemicals in a renewable manner, its progress is still hindered by a lack of predictive capabilities. Media optimization is a critical, and often overlooked, process which is essential to obtain the titers, rates and yields needed for commercial viability. Here, we present a molecule- and host-agnostic active learning process for media optimization that is enabled by a fast and highly repeatable semi-automated pipeline. Its application yielded 60% and 70% increases in titer, and 350% increase in process yield in three different campaigns for flaviolin production in Pseudomonas putida KT2440. Explainable Artificial Intelligence techniques pinpointed that, surprisingly, common salt (NaCl) is the most important component influencing production. The optimal salt concentration is very high, comparable to seawater and close to the limits that P. putida can tolerate. The availability of fast Design-Build-Test-Learn (DBTL) cycles allowed us to show that performance improvements for active learning are rarely monotonous. This work illustrates how machine learning and automation can change the paradigm of current synthetic biology research to make it more effective and informative, and suggests a cost-effective and underexploited strategy to facilitate the high titers, rates and yields essential for commercial viability.
27 July 2024 in Nature Communications
Engineering metabolism to efficiently produce chemicals from multi-step pathways requires optimizing multi-gene expression programs to achieve enzyme balance. CRISPR-Cas transcriptional control systems are emerging as important tools for programming multi-gene expression, but poor predictability of guide RNA folding can disrupt expression control. Here, we correlate efficacy of modified guide RNAs (scRNAs) for CRISPR activation (CRISPRa) in E. coli with a computational kinetic parameter describing scRNA folding rate into the active structure (rS = 0.8). This parameter also enables forward design of scRNAs, allowing us to design a system of three synthetic CRISPRa promoters that can orthogonally activate (>35-fold) expression of chosen outputs. Through combinatorial activation tuning, we profile a three-dimensional design space expressing two different biosynthetic pathways, demonstrating variable production of pteridine and human milk oligosaccharide products. This RNA design approach aids combinatorial optimization of metabolic pathways and may accelerate routine design of effective multi-gene regulation programs in bacterial hosts.
13 December 2025 in Nature Communications
Advances in genome engineering have improved our ability to perturb microbial metabolic networks, yet bioproduction campaigns often struggle with parsing complex metabolic datasets to efficiently enhance product titers. We address this challenge by coupling laboratory automation with machine learning to systematically optimize the production of isoprenol, a sustainable aviation fuel precursor, in Pseudomonas putida. The simultaneous downregulation through CRISPR interference of combinations of up to four gene targets, guided by machine learning, permitted us to increase isoprenol titer 5-fold in six consecutive design-build-test-learn cycles. Moreover, machine learning enabled us to swiftly explore a vast experimental design space of 800,000 possible combinations by strategically recommending approximately 400 priority constructs. High-throughput proteomics allowed us to validate CRISPRi downregulation and identify biological mechanisms driving production increases. Our work demonstrates that ML-driven automated design-build-test-learn cycles, when combined with rigorous data validation, can rapidly enhance titers without specific biological knowledge, suggesting that it can be applied to any host, product, or pathway.
18 April 2024 in npj Systems Biology and Applications
Single cancer cells within a tumor exhibit variable levels of resistance to drugs, ultimately leading to treatment failures. While tumor heterogeneity is recognized as a major obstacle to cancer therapy, standard dose-response measurements for the potency of targeted kinase inhibitors aggregate populations of cells, obscuring intercellular variations in responses. In this work, we develop an analytical and experimental framework to quantify and model dose responses of individual cancer cells to drugs. We first explore the connection between population and single-cell dose responses using a computational model, revealing that multiple heterogeneous populations can yield nearly identical population dose responses. We demonstrate that a single-cell analysis method, which we term a threshold inhibition surface, can differentiate among these populations. To demonstrate the applicability of this method, we develop a dose-titration assay to measure dose responses in single cells. We apply this assay to breast cancer cells responding to phosphatidylinositol-3-kinase inhibition (PI3Ki), using clinically relevant PI3Kis on breast cancer cell lines expressing fluorescent biosensors for kinase activity. We demonstrate that MCF-7 breast cancer cells exhibit heterogeneous dose responses with some cells requiring over ten-fold higher concentrations than the population average to achieve inhibition. Our work reimagines dose-response relationships for cancer drugs in an emerging paradigm of single-cell tumor heterogeneity.
28 October 2025 in PLOS Computational Biology
Cancer metastasis, which requires migration of cancer cells away from the primary tumor, is responsible for approximately 65% percent of cancer-related deaths. Therefore, targeting signaling pathways that drive cancer cell migration or proliferation is a common therapeutic approach. Cell migration is commonly studied using experimental approaches which track cells or cell monolayers as they evolve over time. Computational modeling can then be used to fit partial differential equation (PDE) models to the data, providing mechanistic insights underlying the observed cell motion, including the contribution of various cellular behaviors such as random motion, directed motion, and cell division. A popular experimental technique, the scratch assay, measures the migration and proliferation-driven cell closure of a scratch in a confluent cell monolayer. However, these assays do not disambiguate between different drivers of scratch closure (for instance between cell proliferation and migration to open space). To improve analysis of this technique, we combine scratch assays, video microscopy, and PDE inference to gain quantitative insight to mechanisms of cell migration and proliferation. We capture the evolution of cell density fields over time using live-cell microscopy and automated image processing. Our PDE inference methods involve the use of weak form-based system identification techniques for cell density dynamics modeled with advection-diffusion-reaction systems. We then compare our method with recent modeling work, finding that our model discovery tool automatically identifies similar models including reaction and diffusion terms from a larger set of bases. We demonstrate the application of this framework on 2-dimensional scratch assays subject to the inhibiting effect of trametinib on wound closure and characterize the results in the context of the quantified uncertainty in our inference approach. Our integrated experimental and computational pipeline can be used to rapidly identify and refine models of cell migration in a variety of contexts, enabling the quantitative measurement of the effect of drugs and other perturbations on cell migration and proliferation with uncertainty accounted for.