Salmi, P. | Search bioRxiv

bioRxiv2024

Accurate non-invasive quantification of astaxanthin content using hyperspectral images and machine learning

Commercial cultivation of the microalgae Haematococcus pluvialis to produce natural astaxanthin has gained significant traction due to the high antioxidant capacity of this pigment and its application in foods, feed, cosmetics and nutraceuticals. However, monitoring of astaxanthin content in cultures remains challenging and relies on invasive, time consuming and expensive approaches. In this study, we employed reflectance hyperspectral imaging (HSI) of H. pluvialis suspensions within the visible spectrum, combined with a 1-dimensional convolutional neural network (CNN) to predict the astaxanthin content (g mg-1) as quantified by high-performance liquid chromatography (HPLC). This approach had low average prediction error (5.9%) across a gradient of astaxanthin contents and was only unreliable at very low contents (<0.6 g mg-1). In addition, our machine learning model outperformed single or dual wavelength linear regression models even when the spectral data was obtained with a spectrophotometer coupled with an integrating sphere. Overall, this study proposes the use of HSI in combination with a CNN for precise non-invasive quantification of astaxanthin in cell suspensions.

bioinformatics↗

bioRxiv2022

Single-cell resolution genetic association analysis of heterogeneous bacterial communities by utilizing droplet digital PCR

Microbial communities often respond to environmental challenges, such as the presence of antibiotics, as a whole. Dissecting these community-level effects into separate acting entities requires the identification of organisms that carry functional genes for the observed feature. However, unculturable microbes are abundant in various environments, hence making the identification challenging. Moreover, while at present the development and application of single-cell tools for eukaryotic cells are enhancing, the comparable methodologies applicable for prokaryotic cells are still scarce and have not gained broad and solid status as tools for investigating microbial populations. Here, we present a cultivation-free technique that can be utilized to link functional genes with the carrying bacterial species at single-cell resolution. The developed protocol is relatively simple to use, utilizes commercially available droplet microfluidics devices, does not require toxic reagents, and eliminates invalid signals emerging from extracellular DNA. We validate the methodology by studying the conjugative transfer of antibiotic resistance plasmids in an environment challenged by antibiotics. Furthermore, the method can be customized for any given genetic trait to accurately identify its hosting subpopulation from a heterogeneous and potentially uncultivable bacterial community. ImportanceBacterial systems usually contain numerous different species that may harbor highly similar or identical genes that confer same phenotypic qualities for the community. To decipher the functions of these systems, we report the development of a novel methodology that enables investigating microbial communities at single-cell level. This user-friendly method utilizes droplet digital PCR (ddPCR) to find and identify carriers of specific genes potentially from various microbial sample types. By pinpointing gene carriers, such as those responsible for antibiotic resistance, this method can provide insights to the behavior of microbial communities and gene transfer therein. By strategically combining the use of common methods (ddPCR and amplicon sequencing), the workflow is highly accessible. Thus, it allows also the researchers without a background in single-cell techniques or access to special equipment to adopt the method for producing single-cell data to serve their own research, enabling new research avenues in microbial genetics and ecology.

microbiology↗