Quantification of nisin concentration from fluorescence-based antimicrobial activity assay using Bayesian calibration.

Bacteriocins are ribosomally synthesized peptides with the innate ability to kill or inhibit growth of other bacteria. In recent years, bacteriocins have received increased interest, as their antimicrobial activity enhances food safety and shelf life by combatting pathogens such as Listeria monocytogenes. They also have application potential as an active pharmaceutical compound to combat multidrug-resistant pathogens. As new bacteriocins continue to be discovered, accelerated workflows for screening, identification, and process development have been developed. However, antimicrobial activity measurement is often still limited with regards to quantification and throughput. Here, we present the use of a non-linear calibration model to infer nisin concentrations in cultivation supernatants of Lactococcus lactis ssp. lactis B1629 using readouts of pHluorin2 fluorescence-based antimicrobial activity assays.

Novel calibration design improves knowledge transfer across products for the characterization of pharmaceutical bioprocesses.

Modern machine learning has the potential to fundamentally change the way bioprocesses are developed. In particular, horizontal knowledge transfer methods, which seek to exploit data from historical processes to facilitate process development for a new product, provide an opportunity to rethink current workflows. In this work, we first assess the potential of two knowledge transfer approaches, meta learning and one-hot encoding, in combination with Gaussian process (GP) models. We compare their performance with GPs trained only on data of the new process, that is, local models. Using simulated mammalian cell culture data, we observe that both knowledge transfer approaches exhibit test set errors that are approximately halved compared to those of the local models when two, four, or eight experiments of the new product are used for training. Subsequently, we address the question whether experiments for a new product could be designed more effectively by exploiting existing knowledge. In particular, we suggest to specifically design a few runs for the novel product to calibrate knowledge transfer models, a task that we coin calibration design. We propose a customized objective function to identify a set of calibration design runs, which exploits differences in the process evolution of historical products. In two simulated case studies, we observed that training with calibration designs yields similar test set errors compared to common design of experiments approaches. However, the former requires approximately four times fewer experiments. Overall, the results suggest that process development could be significantly streamlined when systematically carrying knowledge from one product to the next.

Why are different estimates of the effective reproductive number so different? A case study on COVID-19 in Germany.

The effective reproductive number Rt has taken a central role in the scientific, political, and public discussion during the COVID-19 pandemic, with numerous real-time estimates of this quantity routinely published. Disagreement between estimates can be substantial and may lead to confusion among decision-makers and the general public. In this work, we compare different estimates of the national-level effective reproductive number of COVID-19 in Germany in 2020 and 2021. We consider the agreement between estimates from the same method but published at different time points (within-method agreement) as well as retrospective agreement across eight different approaches (between-method agreement). Concerning the former, estimates from some methods are very stable over time and hardly subject to revisions, while others display considerable fluctuations. To evaluate between-method agreement, we reproduce the estimates generated by different groups using a variety of statistical approaches, standardizing analytical choices to assess how they contribute to the observed disagreement. These analytical choices include the data source, data pre-processing, assumed generation time distribution, statistical tuning parameters, and various delay distributions. We find that in practice, these auxiliary choices in the estimation of Rt may affect results at least as strongly as the selection of the statistical approach. They should thus be communicated transparently along with the estimates.

From frozen cell bank to product assay: high-throughput strain characterisation for autonomous Design-Build-Test-Learn cycles.

BACKGROUND: Modern genome editing enables rapid construction of genetic variants, which are further developed in Design-Build-Test-Learn cycles. To operate such cycles in high throughput, fully automated screening, including cultivation and analytics, is crucial in the Test phase. Here, we present the required steps to meet these demands, resulting in an automated microbioreactor platform that facilitates autonomous phenotyping from cryo culture to product assay. RESULTS: First, an automated deep freezer was integrated into the robotic platform to provide working cell banks at all times. A mobile cart allows flexible docking of the freezer to multiple platforms. Next, precultures were integrated within the microtiter plate for cultivation, resulting in highly reproducible main cultures as demonstrated for Corynebacterium glutamicum. To avoid manual exchange of microtiter plates after cultivation, two clean-in-place strategies were established and validated, resulting in restored sterile conditions within two hours. Combined with the previous steps, these changes enable a flexible start of experiments and greatly increase the walk-away time. CONCLUSIONS: Overall, this work demonstrates the capability of our microbioreactor platform to perform autonomous, consecutive cultivation and phenotyping experiments. As highlighted in a case study of cutinase-secreting strains of C. glutamicum, the new procedure allows for flexible experimentation without human interaction while maintaining high reproducibility in early-stage screening processes.

Explore or exploit? A model-based screening strategy for PETase secretion by Corynebacterium glutamicum.

Extracellular production of target proteins simplifies downstream processing due to obsolete cell disruption. However, optimal combinations of a heterologous protein, suitable signal peptide, and secretion host can currently not be predicted, resulting in large strain libraries that need to be tested. On the experimental side, this challenge can be tackled by miniaturization, parallelization, and automation, which provide high-throughput screening data. These data need to be condensed into a candidate ranking for decision-making to focus bioprocess development on the most promising candidates. We screened for Bacillus subtilis signal peptides mediating Sec secretion of two polyethylene terephthalate degrading enzymes (PETases), leaf-branch compost cutinase (LCC) and polyester hydrolase mutants, by Corynebacterium glutamicum. We developed a fully automated screening process and constructed an accompanying Bayesian statistical modeling framework, which we applied in screenings for highest activity in 4-nitrophenyl palmitate degradation. In contrast to classical evaluation methods, batch effects and biological errors are taken into account and their uncertainty is quantified. Within only two rounds of screening, the most suitable signal peptide was identified for each PETase. Results from LCC secretion in microliter-scale cultivation were shown to be scalable to laboratory-scale bioreactors. This work demonstrates an experiment-modeling loop that can accelerate early-stage screening in a way that experimental capacities are focused to the most promising strain candidates. Combined with high-throughput cloning, this paves the way for using large strain libraries of several hundreds of strains in a Design-Build-Test-Learn approach.

Machine learning in bioprocess development: from promise to practice.

Fostered by novel analytical techniques, digitalization, and automation, modern bioprocess development provides large amounts of heterogeneous experimental data, containing valuable process information. In this context, data-driven methods like machine learning (ML) approaches have great potential to rationally explore large design spaces while exploiting experimental facilities most efficiently. Herein we demonstrate how ML methods have been applied so far in bioprocess development, especially in strain engineering and selection, bioprocess optimization, scale-up, monitoring, and control of bioprocesses. For each topic, we will highlight successful application cases, current challenges, and point out domains that can potentially benefit from technology transfer and further progress in the field of ML.

Accelerated strain construction and characterization of C. glutamicum protein secretion by laboratory automation.

Secretion of bacterial proteins into the culture medium simplifies downstream processing by avoiding cell disruption for target protein purification. However, a suitable signal peptide for efficient secretion needs to be identified, and currently, there are no tools available to predict optimal combinations of signal peptides and target proteins. The selection of such a combination is influenced by several factors, including protein biosynthesis efficiency and cultivation conditions, which both can have a significant impact on secretion performance. As a result, a large number of combinations must be tested. Therefore, we have developed automated workflows allowing for targeted strain construction and secretion screening using two platforms. Key advantages of this experimental setup include lowered hands-on time and increased throughput. In this study, the automated workflows were established for the heterologous production of Fusarium solani f. sp. pisi cutinase in Corynebacterium glutamicum. The target protein was monitored in culture supernatants via enzymatic activity and split GFP assay. Varying spacer lengths between the Shine-Dalgarno sequence and the start codon of Bacillus subtilis signal peptides were tested. Consistent with previous work on the secretory cutinase production in B. subtilis, a ribosome binding site with extended spacer length to up to 12 nt, which likely slows down translation initiation, does not necessarily lead to poorer cutinase secretion by C. glutamicum. The best performing signal peptides for cutinase secretion with a standard spacer length were identified in a signal peptide screening. Additional insights into the secretion process were gained by monitoring secretion stress using the C. glutamicum K9 biosensor strain. KEY POINTS: • Automated workflows for strain construction and screening of protein secretion • Comparison of spacer, signal peptide, and host combinations for cutinase secretion • Signal peptide screening for secretion by C. glutamicum using the split GFP assay.