Chemistry-Based Modeling on Phenotype-Based Drug-Induced Liver Injury Annotation: From Public to Proprietary Data.

Drug-induced liver injury (DILI) is an important safety concern and a major reason to remove a drug from the market. Advancements in recent machine learning methods have led to a wide range of in silico models for DILI predictive methods based on molecule chemical structures (fingerprints). Existing publicly available DILI data sets used for model building are based on the interpretation of drug labels or patient case reports, resulting in a typical binary clinical DILI annotation. We developed a novel phenotype-based annotation to process hepatotoxicity information extracted from repeated dose in vivo preclinical toxicology studies using INHAND annotation to provide a more informative and reliable data set for machine learning algorithms. This work resulted in a data set of 430 unique compounds covering diverse liver pathology findings which were utilized to develop multiple DILI prediction models trained on the publicly available data (TG-GATEs) using the compound's fingerprint. We demonstrate that the TG-GATEs compounds DILI labels can be predicted well and how the differences between TG-GATEs and the external test compounds (Johnson & Johnson) impact the model generalization performance.

Leveraging Cell Painting Images to Expand the Applicability Domain and Actively Improve Deep Learning Quantitative Structure-Activity Relationship Models.

The search for chemical hit material is a lengthy and increasingly expensive drug discovery process. To improve it, ligand-based quantitative structure-activity relationship models have been broadly applied to optimize primary and secondary compound properties. Although these models can be deployed as early as the stage of molecule design, they have a limited applicability domain─if the structures of interest differ substantially from the chemical space on which the model was trained, a reliable prediction will not be possible. Image-informed ligand-based models partly solve this shortcoming by focusing on the phenotype of a cell caused by small molecules, rather than on their structure. While this enables chemical diversity expansion, it limits the application to compounds physically available and imaged. Here, we employ an active learning approach to capitalize on both of these methods' strengths and boost the model performance of a mitochondrial toxicity assay (Glu/Gal). Specifically, we used a phenotypic Cell Painting screen to build a chemistry-independent model and adopted the results as the main factor in selecting compounds for experimental testing. With the additional Glu/Gal annotation for selected compounds we were able to dramatically improve the chemistry-informed ligand-based model with respect to the increased recognition of compounds from a 10% broader chemical space.

Using single-plant-omics in the field to link maize genes to functions and phenotypes.

Most of our current knowledge on plant molecular biology is based on experiments in controlled laboratory environments. However, translating this knowledge from the laboratory to the field is often not straightforward, in part because field growth conditions are very different from laboratory conditions. Here, we test a new experimental design to unravel the molecular wiring of plants and study gene-phenotype relationships directly in the field. We molecularly profiled a set of individual maize plants of the same inbred background grown in the same field and used the resulting data to predict the phenotypes of individual plants and the function of maize genes. We show that the field transcriptomes of individual plants contain as much information on maize gene function as traditional laboratory-generated transcriptomes of pooled plant samples subject to controlled perturbations. Moreover, we show that field-generated transcriptome and metabolome data can be used to quantitatively predict individual plant phenotypes. Our results show that profiling individual plants in the field is a promising experimental design that could help narrow the lab-field gap.

Leaf Growth Response to Mild Drought: Natural Variation in Arabidopsis Sheds Light on Trait Architecture.

Plant growth and crop yield are negatively affected by a reduction in water availability. However, a clear understanding of how growth is regulated under nonlethal drought conditions is lacking. Recent advances in genomics, phenomics, and transcriptomics allow in-depth analysis of natural variation. In this study, we conducted a detailed screening of leaf growth responses to mild drought in a worldwide collection of Arabidopsis thaliana accessions. The genetic architecture of the growth responses upon mild drought was investigated by subjecting the different leaf growth phenotypes to genome-wide association mapping and by characterizing the transcriptome of young developing leaves. Although no major effect locus was found to be associated with growth in mild drought, the transcriptome analysis delivered further insight into the natural variation of transcriptional responses to mild drought in a specific tissue. Coexpression analysis indicated the presence of gene clusters that co-vary over different genetic backgrounds, among others a cluster of genes with important regulatory functions in the growth response to osmotic stress. It was found that the occurrence of a mild drought stress response in leaves can be inferred with high accuracy across accessions based on the expression profile of 283 genes. A genome-wide association study on the expression data revealed that trans regulation seems to be more important than cis regulation in the transcriptional response to environmental perturbations.

The reduction in maize leaf growth under mild drought affects the transition between cell division and cell expansion and cannot be restored by elevated gibberellic acid levels.

Growth is characterized by the interplay between cell division and cell expansion, two processes that occur separated along the growth zone at the maize leaf. To gain further insight into the transition between cell division and cell expansion, conditions were investigated in which the position of this transition zone was positively or negatively affected. High levels of gibberellic acid (GA) in plants overexpressing the GA biosynthesis gene GA20-OXIDASE (GA20OX-1OE ) shifted the transition zone more distally, whereas mild drought, which is associated with lowered GA biosynthesis, resulted in a more basal positioning. However, the increased levels of GA in the GA20OX-1OE line were insufficient to convey tolerance to the mild drought treatment, indicating that another mechanism in addition to lowered GA levels is restricting growth during drought. Transcriptome analysis with high spatial resolution indicated that mild drought specifically induces a reprogramming of transcriptional regulation in the division zone. 'Leaf Growth Viewer' was developed as an online searchable tool containing the high-resolution data.

Natural Variation of Molecular and Morphological Gibberellin Responses.

Although phytohormones such as gibberellins are essential for many conserved aspects of plant physiology and development, plants vary greatly in their responses to these regulatory compounds. Here, we use genetic perturbation of endogenous gibberellin levels to probe the extent of intraspecific variation in gibberellin responses in natural accessions of Arabidopsis (Arabidopsis thaliana). We find that these accessions vary greatly in their ability to buffer the effects of overexpression of GA20ox1, encoding a rate-limiting enzyme for gibberellin biosynthesis, with substantial differences in bioactive gibberellin concentrations as well as transcriptomes and growth trajectories. These findings demonstrate a surprising level of flexibility in the wiring of regulatory networks underlying hormone metabolism and signaling.

Combined Large-Scale Phenotyping and Transcriptomics in Maize Reveals a Robust Growth Regulatory Network.

Leaves are vital organs for biomass and seed production because of their role in the generation of metabolic energy and organic compounds. A better understanding of the molecular networks underlying leaf development is crucial to sustain global requirements for food and renewable energy. Here, we combined transcriptome profiling of proliferative leaf tissue with in-depth phenotyping of the fourth leaf at later stages of development in 197 recombinant inbred lines of two different maize (Zea mays) populations. Previously, correlation analysis in a classical biparental mapping population identified 1,740 genes correlated with at least one of 14 traits. Here, we extended these results with data from a multiparent advanced generation intercross population. As expected, the phenotypic variability was found to be larger in the latter population than in the biparental population, although general conclusions on the correlations among the traits are comparable. Data integration from the two diverse populations allowed us to identify a set of 226 genes that are robustly associated with diverse leaf traits. This set of genes is enriched for transcriptional regulators and genes involved in protein synthesis and cell wall metabolism. In order to investigate the molecular network context of the candidate gene set, we integrated our data with publicly available functional genomics data and identified a growth regulatory network of 185 genes. Our results illustrate the power of combining in-depth phenotyping with transcriptomics in mapping populations to dissect the genetic control of complex traits and present a set of candidate genes for use in biomass improvement.

Low concentration cell painting images enable the identification of highly potent compounds.

Image-based models that use features extracted from cell microscopy images can estimate the activity of small molecules in various biological assays. Typically, models are trained on images stained by an optimized protocol (e.g. Cell Painting) after exposure to a fairly high small molecule concentration (referred to as 'image concentration') of 10 µ M or higher. Low concentration images (e.g. 0.16 µM, 0.8 µM, 4 µM) tend to yield models with worse performance. In this work, we nevertheless report a practical use for low image concentration data. We propose the combination of well-performing models trained at higher image concentrations, with lower image concentration for inference to identify more potent compounds. We show that this approach improves on the conventional method (directly training a high-potency model) in 65 % of assays investigated in terms of AUC-ROC, and 75 % of assays in terms of RIPtoP-corrected AUC-PR.

Selection of organisms for the co-evolution-based study of protein interactions.

BACKGROUND: The prediction and study of protein interactions and functional relationships based on similarity of phylogenetic trees, exemplified by the mirrortree and related methodologies, is being widely used. Although dependence between the performance of these methods and the set of organisms used to build the trees was suspected, so far nobody assessed it in an exhaustive way, and, in general, previous works used as many organisms as possible. In this work we asses the effect of using different sets of organism (chosen according with various phylogenetic criteria) on the performance of this methodology in detecting protein interactions of different nature. RESULTS: We show that the performance of three mirrortree-related methodologies depends on the set of organisms used for building the trees, and it is not always directly related to the number of organisms in a simple way. Certain subsets of organisms seem to be more suitable for the predictions of certain types of interactions. This relationship between type of interaction and optimal set of organism for detecting them makes sense in the light of the phylogenetic distribution of the organisms and the nature of the interactions. CONCLUSIONS: In order to obtain an optimal performance when predicting protein interactions, it is recommended to use different sets of organisms depending on the available computational resources and data, as well as the type of interactions of interest.

Correlation analysis of the transcriptome of growing leaves with mature leaf parameters in a maize RIL population.

BACKGROUND: To sustain the global requirements for food and renewable resources, unraveling the molecular networks underlying plant growth is becoming pivotal. Although several approaches to identify genes and networks involved in final organ size have been proven successful, our understanding remains fragmentary. RESULTS: Here, we assessed variation in 103 lines of the Zea mays B73xH99 RIL population for a set of final leaf size and whole shoot traits at the seedling stage, complemented with measurements capturing growth dynamics, and cellular measurements. Most traits correlated well with the size of the division zone, implying that the molecular basis of final leaf size is already defined in dividing cells of growing leaves. Therefore, we searched for association between the transcriptional variation in dividing cells of the growing leaf and final leaf size and seedling biomass, allowing us to identify genes and processes correlated with the specific traits. A number of these genes have a known function in leaf development. Additionally, we illustrated that two independent mechanisms contribute to final leaf size, maximal growth rate and the duration of growth. CONCLUSIONS: Untangling complex traits such as leaf size by applying in-depth phenotyping allows us to define the relative contributions of the components and their mutual associations, facilitating dissection of the biological processes and regulatory networks underneath.

Adaptation for protein synthesis efficiency in a naturally occurring self-regulating operon.

The korAB operon in RK2 plasmids is a beautiful natural example of a negatively and cooperatively self-regulating operon. It has been particularly well characterized both experimentally and with mathematical models. We have carried out a detailed investigation of the role of the regulatory mechanism using a biologically grounded mechanistic multi-scale stochastic model that includes plasmid gene regulation and replication in the context of host growth and cell division. We use the model to compare four hypotheses for the action of the regulatory mechanism: increased robustness to extrinsic factors, decreased protein fluctuations, faster response-time of the operon and reduced host burden through improved efficiency of protein production. We find that the strongest impact of all elements of the regulatory architecture is on improving the efficiency of protein synthesis by reduction in the number of mRNA molecules needed to be produced, leading to a greater than ten-fold reduction in host energy required to express these plasmid proteins. A smaller but still significant role is seen for speeding response times, but this is not materially improved by the cooperativity. The self-regulating mechanisms have the least impact on protein fluctuations and robustness. While reduction of host burden is evident in a plasmid context, negative self-regulation is a widely seen motif for chromosomal genes. We propose that an important evolutionary driver for negatively self-regulated genes is to improve the efficiency of protein synthesis.

Global transcription regulation of RK2 plasmids: a case study in the combined use of dynamical mathematical models and statistical inference for integration of experimental data and hypothesis exploration.

BACKGROUND: IncP-1 plasmids are broad host range plasmids that have been found in clinical and environmental bacteria. They often carry genes for antibiotic resistance or catabolic pathways. The archetypal IncP-1 plasmid RK2 is a well-characterized biological system, with a fully sequenced and annotated genome and wide range of experimental measurements. Its central control operon, encoding two global regulators KorA and KorB, is a natural example of a negatively self-regulated operon. To increase our understanding of the regulation of this operon, we have constructed a dynamical mathematical model using Ordinary Differential Equations, and employed a Bayesian inference scheme, Markov Chain Monte Carlo (MCMC) using the Metropolis-Hastings algorithm, as a way of integrating experimental measurements and a priori knowledge. We also compared MCMC and Metabolic Control Analysis (MCA) as approaches for determining the sensitivity of model parameters. RESULTS: We identified two distinct sets of parameter values, with different biological interpretations, that fit and explain the experimental data. This allowed us to highlight the proportion of repressor protein as dimers as a key experimental measurement defining the dynamics of the system. Analysis of joint posterior distributions led to the identification of correlations between parameters for protein synthesis and partial repression by KorA or KorB dimers, indicating the necessary use of joint posteriors for correct parameter estimation. Using MCA, we demonstrated that the system is highly sensitive to the growth rate but insensitive to repressor monomerization rates in their selected value regions; the latter outcome was also confirmed by MCMC. Finally, by examining a series of different model refinements for partial repression by KorA or KorB dimers alone, we showed that a model including partial repression by KorA and KorB was most compatible with existing experimental data. CONCLUSIONS: We have demonstrated that the combination of dynamical mathematical models with Bayesian inference is valuable in integrating diverse experimental data and identifying key determinants and parameters for the IncP-1 central control operon. Moreover, we have shown that Bayesian inference and MCA are complementary methods for identification of sensitive parameters. We propose that this demonstrates generic value in applying this combination of approaches to systems biology dynamical modelling.