Natural compounds constitute a rich resource of potential small molecule therapeutics. While experimental access to this resource is limited due to its vast diversity and difficulties in systematic purification, computational assessment of structural similarity with known therapeutic molecules offers a scalable approach. Here, we assessed functional similarity between natural compounds and approved drugs by combining multiple chemical similarity metrics and physicochemical properties using a machine-learning approach. We computed pairwise similarities between 1410 drugs for training classification models and used the drugs shared protein targets as class labels. The best performing models were random forest which gave an average area under the ROC of 0.9, Matthews correlation coefficient of 0.35, and F1 score of 0.33, suggesting that it captured the structure-activity relation well. The models were then used to predict protein targets of circa 11k natural compounds by comparing them with the drugs. This revealed therapeutic potential of several natural compounds, including those with support from previously published sources as well as those hitherto unexplored. We experimentally validated one of the predicted pair’s activities, viz., Cox-1 inhibition by 5-methoxysalicylic acid, a molecule commonly found in tea, herbs and spices. In contrast, another natural compound, 4-isopropylbenzoic acid, with the highest similarity score when considering most weighted similarity metric but not picked by our models, did not inhibit Cox-1. Our results demonstrate the utility of a machine-learning approach combining multiple chemical features for uncovering protein binding potential of natural compounds.
Cross‐feeding is fundamental to the diversity and function of microbial communities. However, identification of cross‐fed metabolites is often challenging due to the universality of metabolic and biosynthetic intermediates. Here, we use 13 C isotope tracing in peptides to elucidate cross‐fed metabolites in co‐cultures of Saccharomyces cerevisiae and Lactococcus lactis . The community was grown on lactose as the main carbon source with either glucose or galactose fraction of the molecule labelled with 13 C. Data analysis allowing for the possible mass‐shifts yielded hundreds of peptides for which we could assign both species identity and labelling degree. The labelling pattern showed that the yeast utilized galactose and, to a lesser extent, lactic acid shared by L. lactis as carbon sources. While the yeast provided essential amino acids to the bacterium as expected, the data also uncovered a complex pattern of amino acid exchange. The identity of the cross‐fed metabolites was further supported by metabolite labelling in the co‐culture supernatant, and by diminished fitness of a galactose‐negative yeast mutant in the community. Together, our results demonstrate the utility of 13 C‐based proteomics for uncovering microbial interactions.
Genetic screens using high-throughput fluorescent microscopes have generated large datasets, contributing many cell biological insights. Such approaches cannot tackle questions requiring knowledge of ultrastructure below the resolution limit of fluorescent microscopy. Electron microscopy (EM) reveals detailed cellular ultrastructure but requires time-consuming sample preparation, limiting throughput. Here we describe a robust method for screening by high-throughput EM. Our approach uses combinations of fluorophores as barcodes to uniquely mark each cell type in mixed populations and correlative light and EM (CLEM) to read the barcode of each cell before it is imaged by EM. Coupled with an easy-to-use software workflow for correlation, segmentation, and computer image analysis, our method, called “MultiCLEM,” allows us to extract and analyze multiple cell populations from each EM sample preparation. We demonstrate several uses for MultiCLEM with 15 different yeast variants. The methodology is not restricted to yeast, can be scaled to higher throughput, and can be used in multiple ways to enable EM to become a powerful screening technique.
A growing number of studies indicate that mRNAs and long ncRNAs can affect protein populations by assembling dynamic ribonucleoprotein (RNP) granules. These phase-separated molecular ‘sponges’, stabilized by quinary (transient and weak) interactions, control proteins involved in numerous biological functions. Retroviruses such as HIV-1 form by self-assembly when their genomic RNA (gRNA) traps Gag and GagPol polyprotein precursors. Infectivity requires extracellular budding of the particle followed by maturation, an ordered processing of ∼2400 Gag and ∼120 GagPol by the viral protease (PR). This leads to a condensed gRNA-NCp7 nucleocapsid and a CAp24-self-assembled capsid surrounding the RNP. The choreography by which all of these components dynamically interact during virus maturation is one of the missing milestones to fully depict the HIV life cycle. Here, we describe how HIV-1 has evolved a dynamic RNP granule with successive weak–strong–moderate quinary NC-gRNA networks during the sequential processing of the GagNC domain. We also reveal two palindromic RNA-binding triads on NC, KxxFxxQ and QxxFxxK, that provide quinary NC-gRNA interactions. Consequently, the nucleocapsid complex appears properly aggregated for capsid reassembly and reverse transcription, mandatory processes for viral infectivity. We show that PR is sequestered within this RNP and drives its maturation/condensation within minutes, this process being most effective at the end of budding. We anticipate such findings will stimulate further investigations of quinary interactions and emergent mechanisms in crowded environments throughout the wide and growing array of RNP granules.
Genetic screens performed using high-throughput fluorescent microscopes have generated large datasets that have contributed many insights into cell biology. However, such approaches typically cannot tackle questions requiring knowledge of ultrastructure below the resolution limit of fluorescent microscopy. Electron microscopy (EM) is not subject to this resolution limit, generating detailed images of cellular ultrastructure, but requires time consuming preparation of individual samples, limiting its throughput. Here we overcome this obstacle and describe a robust method for screening by high-throughput electron microscopy. Our approach uses combinations of fluorophores as barcodes to mark the genotype of each cell in mixed populations, and correlative light and electron microscopy to read the fluorescent barcode of each cell before it is imaged by electron microscopy. Coupled with an easy-to-use software workflow for correlation, segmentation and computer image analysis, our method allows to extract and analyze multiple cell populations from each EM sample preparation. We demonstrate the method on several organelles with samples that each contain up to 15 different yeast variants. The methodology is not restricted to yeast, can be scaled to higher-throughput, and can be utilized in multiple ways to enable electron microscopy to become a powerful screening methodology.