Omics Measurements Research Articles

DeePathNet: a transformer-based deep learning model integrating multi-omic data with cancer pathways.

Multi-omic data analysis incorporating machine learning has the potential to significantly improve cancer diagnosis and prognosis. Traditional machine learning methods are usually limited to omic measurements, omitting existing domain knowledge, such as the biological networks that link molecular entities in various omic data types. Here we develop a Transformer-based explainable deep learning model, DeePathNet, which integrates cancer-specific pathway information into multi-omic data analysis. Using a variety of big datasets, including ProCan-DepMapSanger, CCLE, and TCGA, we demonstrate and validate that DeePathNet outperforms traditional methods for predicting drug response and classifying cancer type and subtype. Combining biomedical knowledge and state-of-the-art deep learning methods, DeePathNet enables biomarker discovery at the pathway level, maximizing the power of data-driven approaches to cancer research. DeePathNet is available on GitHub at https://github.com/CMRI-ProCan/DeePathNet.

The Addition of Transcriptomics to the Bead-Enabled Accelerated Monophasic Multi-Omics Method: A Step toward Universal Sample Preparation.

Omics-based measurements enable the study of biomolecules in a high-throughput fashion, leading to the characterization and quantification of biological systems. Multi-omics methods aim to incorporate several omics measurements for a more holistic approach, which is crucial for advancing our understanding of the diversity and redundancy of biological systems. Current multi-omics sample preparation methods have achieved proteomics, lipidomics, and metabolomics from individual samples; however, the bioinformatic tools currently available for interpreting data generated from these omics are limited. Alternately, transcriptomics has a wide arsenal of available bioinformatic tools offering intensive sample characterization but has yet to be incorporated into a unified, multi-omics sample preparation technique. Herein we describe the modified bead-enabled accelerated monophasic multi-omics (mBAMM) method, which incorporates RNA extraction for transcriptomics analysis. mBAMM was shown to enable RNA-seq without compromising the isolation of biomolecules for proteomics, lipidomics, and metabolomics. This methodology greatly improves sample characterization and represents a major innovation toward cohesive insights into biological systems.

Bootstrap Evaluation of Association Matrices (BEAM) for Integrating Multiple Omics Profiles with Multiple Outcomes.

Large datasets containing multiple clinical and omics measurements for each subject motivate the development of new statistical methods to integrate these data to advance scientific discovery. We propose bootstrap evaluation of association matrices (BEAM), which integrates multiple omics profiles with multiple clinical endpoints. BEAM associates a set omic features with clinical endpoints via regression models and then uses bootstrap resampling to determine statistical significance of the set. Unlike existing methods, BEAM uniquely accommodates an arbitrary number of omic profiles and endpoints. In simulations, BEAM performed similarly to the theoretically best simple test and outperformed other integrated analysis methods. In an example pediatric leukemia application, BEAM identified several genes with biological relevance established by a CRISPR assay that had been missed by univariate screens and other integrated analysis methods. Thus, BEAM is a powerful, flexible, and robust tool to identify genes for further laboratory and/or clinical research evaluation. Source code, documentation, and a vignette for BEAM are available on GitHub at: https://github.com/annaSeffernick/BEAMR . The R package is available from CRAN at: https://cran.r-project.org/package=BEAMR . Stanley.Pounds@stjude.org. Supplementary data are available at the journal's website.

Deciphering spatial domains from spatial multi-omics with SpatialGlue

Advances in spatial omics technologies now allow multiple types of data to be acquired from the same tissue slice. To realize the full potential of such data, we need spatially informed methods for data integration. Here, we introduce SpatialGlue, a graph neural network model with a dual-attention mechanism that deciphers spatial domains by intra-omics integration of spatial location and omics measurement followed by cross-omics integration. We demonstrated SpatialGlue on data acquired from different tissue types using different technologies, including spatial epigenome–transcriptome and transcriptome–proteome modalities. Compared to other methods, SpatialGlue captured more anatomical details and more accurately resolved spatial domains such as the cortex layers of the brain. Our method also identified cell types like spleen macrophage subsets located at three different zones that were not available in the original data annotations. SpatialGlue scales well with data size and can be used to integrate three modalities. Our spatial multi-omics analysis tool combines the information from complementary omics modalities to obtain a holistic view of cellular and tissue properties.

#894 Dynamic multi-omics and mechanistic modelling approach provides deeper insight into kidney fibrosis progression

Abstract Background and Aims Chronic kidney disease (CKD) affects more than 10% of the world's population and causes millions of deaths annually. Organ fibrosis is the main driver of CKD, leading to excessive accumulation of extracellular matrix (ECM). The presence of myofibroblasts correlates with disease progression. The pro-fibrotic responses of these cells, such as ECM production, are driven by the master regulator TGF-β. Despite progress in understanding the disease, there are no treatments available, and current diagnostic indicators neither facilitate early detection of CKD nor correlate with actual renal damage [1–4] Using human kidney-derived PDGFRβ+ mesenchymal cells we aim to provide time-resolved mechanistic insight into CKD progression. Method Our goal is to understand and model how these cells contribute to fibrosis by studying their kinetic response to TGF-β. Therefore, we use multiple omics measurements: transcriptomics, secretomics and phospho-/proteomics coupled with quantitative imaging of the ECM. For each time point, we quantified differential changes in omics and estimated the activities of transcription factors, kinases and phosphatases [5]. Furthermore, we used causal reasoning to draw mechanistic relationships between omics layers within and across time points. Results For validation, we focused on a set of transcription factors that also showed importance in the network context that considers several data modalities. Using knock-down experiments, we show the relevance of these transcription factors for the ECM deposition as well as the fibrotic response to TGF-β. Conclusion Taken together, by applying mechanistic modeling and the integration of multiple datasets, we are contributing to a better understanding of the factors and mechanisms driving kidney fibrosis. Furthermore, we will publish this dataset and thus provide a valuable resource for researchers in the field of CKD.

Preclinical side effect prediction through pathway engineering of protein interaction network models.

Modeling tools aim to predict potential drug side effects, although they suffer from imperfect performance. Specifically, protein-protein interaction models predict drug effects from proteins surrounding drug targets, but they tend to overpredict drug phenotypes and require well-defined pathway phenotypes. In this study, we used PathFX, a protein-protein interaction tool, to predict side effects for active ingredient-side effect pairs extracted from drug labels. We observed limited performance and defined new pathway phenotypes using pathway engineering strategies. We defined new pathway phenotypes using a network-based and gene expression-based approach. Overall, we discovered a trade-off between sensitivity and specificity values and demonstrated a way to limit overprediction for side effects with sufficient true positive examples. We compared our predictions to animal models and demonstrated similar performance metrics, suggesting that protein-protein interaction models do not need perfect evaluation metrics to be useful. Pathway engineering, through the inclusion of true positive examples and omics measurements, emerges as a promising approach to enhance the utility of protein interaction network models for drug effect prediction.

Abstract 4869: Efficient study design for the discovery of a gene expression signature predicting metastasis in cutaneous squamous cell carcinoma

Abstract Background: The StepIdent study aims to develop a gene signature predicting metastasis in patients with cutaneous squamous cell carcinoma (cSCC) to improve risk stratification, thus enabling personalized decisions about follow-up schedules and treatment options. Here we describe the unique characteristics, challenges, and best practices for an efficient design of a discovery cohort for a rare outcome (metastasis prevalence: 2-5%); for retrieving, curating, and linking the clinical and pathological data through nationwide databases; and for measuring gene expression through sequencing of archived Formalin-Fixed Paraffin-Embedded (FFPE) primary tumor samples. Methods: Following a predefined protocol, we identified a nested-case control cohort (NCC) of 305 cases and 305 controls from a nationwide cohort of 19,120 patients with a first cSCC in the Netherlands from 2007 to 2009, followed up until 2020. We chose an NCC design since it is an efficient study design in a rare outcome setting (weighting is needed to accommodate the under-sampling of the controls). Patients were identified from the Dutch National Cancer Registry (NCR) and the clinical information was retrieved from the NCR which is linked to the nationwide registry of histo- and cytopathology (PALGA). Tumor blocks were requested from PALGA, and pathological characteristics were assessed by dermatopathologists. We matched controls to cases, based on a risk score estimated by a clinicopathological model. Gene expression was measured using the Illumina RNA Prep with Enrichment kit combined with the whole exome panel and paired-end sequenced on the NextSeq 550. Results: Tissue slides for 541 samples were retrieved for sequencing. 151 samples were excluded after pathology review or due to low pre-library concentration. The final cohort includes 195 case-control pairs (n=390). The median sequencing depth was 43M (Q1-Q3: 35-52M); the median Q30 was 85% (Q1-Q3: 83-87%); the median GC content was 51% (Q1-Q3: 50-52%); a median of 1.8% of base pairs (Q1-Q3: 1.4-2.1%) was trimmed prior to the mapping/alignment; a median of 69% (Q1-Q3: 65-74%) of reads were aligned as protein-coding and a median of 7% (Q1-Q3: 6-10%) as rRNA; a median of 95% (Q1-Q3: 93-96%) of reads were aligned by STAR. Two samples were excluded based on quality control. Conclusion: We described an efficient design and implementation of a nationwide discovery study in cSCC, involving the retrieval of clinicopathological data, the collection of FFPE materials, and the execution of omics measurements. This study presents the largest cohort to date, incorporating omics measurements of primary cSCC samples, combined with simultaneous access to well-curated clinical and pathological information and follow-up data. Our findings can provide guidance for similar studies involving a rare clinical endpoint, where an efficient study design is a necessity. Citation Format: Barbara Rentroia-Pacheco, Lara Pozza, Yan Ting Chen, Daphne Huigh, Celeste J. Eggermont, Olivia FM Steijlen, Sheril Alex, Jvalini Dwarkasing, Domenico Bellomo, Harmen JG van de Werken, Antien L. Mooyaart, Marlies Wakkee, Loes M. Hollestein. Efficient study design for the discovery of a gene expression signature predicting metastasis in cutaneous squamous cell carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4869.

Abstract 7393: Tumor model to tumor treatment: Applying deep learning approaches to map multimodal data from cancer model systems to patients

Abstract Despite large-scale efforts to measure the effect of drug screens in cancer cell lines, mapping the effects of drugs to patient samples has been a challenge. Biological differences between cell lines and patients, such as lack of immune system or microbiome, in-vitro survival adaptations, and biases in measurement technologies create differences across sample modalities that can confound analysis including prediction with machine learning. In this work, we propose a multiway batch correction strategy to enable algorithmic prediction of tumor drug response across model systems and patient data.Recent advances in batch correction algorithms have been motivated by the need to correct for batch effects in single-cell omics and include diverse approaches such as variational autoencoders (VAEs) and generative adversarial networks (GANs). Given the successes of these generative deep learning methods in single cell sequencing analysis, we worked to employ similar approaches to correct large omics measurements across various cancer datasets. Here, we describe mapping of datasets from diverse data sources and model systems to the same space, so that a predictive model of drug response built in a system such as cell lines can be used in biologically relevant models such as organoids, patient derived xenografts, and tumor data. Specifically, we introduce a modified loss function in a VAE using cosine similarity distance to minimize the effect of different cancer model systems in predicting cancer types. We evaluate the method on standard data types for drug response prediction - gene expression, copy number variation, and protein abundance. For this method, the cosine similarity is added as an additional term to the VAE reconstruction and Kullback-Leibler divergence loss terms. This injects a quantification of the dissimilarity between the tumor and tumor model distributions into the backpropagation and gradient descent for updating the model parameters resulting in an encoded representation of the data where the effect of data source has been attenuated while preserving the phenotypic signal. We evaluate our approach for biological signal preservation while reducing model system-specific noise with logistic regression and Euclidean distance. Our results show that the proposed VAE can effectively correct for platform effects and improve the accuracy of downstream integrative analyses. This study has the potential to improve the accuracy and translatability of proteogenomic drug response studies. The proposed modified VAE could be used to correct for platform effects in a variety of datasets, including those from different studies, different platforms, and different cancer types. This could lead to new insights into cancer biology, calibration of cancer patient digital twins, and the development of new diagnostic and therapeutic strategies. Citation Format: Brian Karlberg, Raphael Kirchgaessner, Jeremy R. Jacobson, Kyle Ellrott, Sara J. Gosline. Tumor model to tumor treatment: Applying deep learning approaches to map multimodal data from cancer model systems to patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7393.

Benchmarking feature selection and feature extraction methods to improve the performances of machine-learning algorithms for patient classification using metabolomics biomedical data

ObjectiveClassification tasks are an open challenge in the field of biomedicine. While several machine-learning techniques exist to accomplish this objective, several peculiarities associated with biomedical data, especially when it comes to omics measurements, prevent their use or good performance achievements. Omics approaches aim to understand a complex biological system through systematic analysis of its content at the molecular level. On the other hand, omics data are heterogeneous, sparse and affected by the classical “curse of dimensionality” problem, i.e. having much fewer observation, samples (n) than omics features (p). Furthermore, a major problem with multi-omics data is the imbalance either at the class or feature level. The objective of this work is to study whether feature extraction and/or feature selection techniques can improve the performances of classification machine-learning algorithms on omics measurements. Methods:Among all omics, metabolomics has emerged as a powerful tool in cancer research, facilitating a deeper understanding of the complex metabolic landscape associated with tumorigenesis and tumor progression. Thus, we selected three publicly available metabolomics datasets, and we applied several feature extraction techniques both linear and non-linear, coupled or not with feature selection methods, and evaluated the performances regarding patient classification in the different configurations for the three datasets. Results:We provide general workflow and guidelines on when to use those techniques depending on the characteristics of the data available. To further test the extension of our approach to other omics data, we have included a transcriptomics and a proteomics data. Overall, for all datasets, we showed that applying supervised feature selection improves the performances of feature extraction methods for classification purposes. Scripts used to perform all analyses are available at: https://github.com/Plant-Net/Metabolomic_project/.

Functional and multi-omics signatures of mitapivat efficacy upon activation of pyruvate kinase in red blood cells from patients with sickle cell disease.

Mitapivat, a pyruvate kinase activator, shows great potential as a sickle cell disease (SCD)-modifying therapy. The safety and efficacy of mitapivat as a long-term maintenance therapy are currently being evaluated in two open-label studies. Here we applied a comprehensive multi-omics approach to investigate the impact of activating pyruvate kinase on red blood cells (RBC) from 15 SCD patients. HbSS patients were enrolled in one of the open-label, extended studies (NCT04610866). Leukodepleted RBC obtained from fresh whole blood at baseline, prior to drug initiation, and at longitudinal timepoints over the course of the study were processed for multi-omics through a stepwise extraction of metabolites, lipids and proteins. Mitapivat therapy had significant effects on the metabolome, lipidome and proteome of SCD RBC. Mitapivat decreased 2,3-diphosphoglycerate levels, increased adenosine triphosphate levels, and improved hematologic and sickling parameters in patients with SCD. Agreement between omics measurements and clinical measurements confirmed the specificity of mitapivat on targeting late glycolysis, with glycolytic metabolites ranking as the top correlates to parameters of hemoglobin S oxygen affinity (p50) and sickling kinetics (t50) during treatment. Mitapivat markedly reduced levels of proteins of mitochondrial origin within 2 weeks of initiation of treatment, with minimal changes in reticulocyte counts. In the first 6 months of treatment there were also transient elevations of lysophosphatidylcholines and oxylipins with depletion of free fatty acids, suggestive of an effect on membrane lipid remodeling. Multi-omics analysis of RBC identified benefits for glycolysis, as well as activation of the Lands cycle.

The Effect of 8-Week Protein Supplementation with a Simple Exercise Program on Body Composition, Muscle Strength, and Amino Acid OMICS among Healthy Sedentary Indians: A Randomized, Double-Blind, Placebo-Controlled Trial.

Dietary protein plays a crucial role in the modulation of several physiological processes to sustain health and well-being. There is robust evidence of enhanced muscle protein synthesis, improved physical fitness, body composition, and performance contributed by protein supplementation combined with exercise among trained individuals or athletes. Evidence of the efficacy of such intervention on healthy adults having a sedentary lifestyle is limited. The objective of this study was to evaluate the impact of 12 g of additional protein in the form of a protein supplement compared to a placebo combined with a simple exercise program on plasma amino acid level, body composition, and muscle strength among healthy Indian adults having a relatively sedentary lifestyle. This double-blinded, randomized controlled trial was conducted on sedentary healthy adults 20-45 years of age, with a body mass index (BMI) between 18.5 and 27.9 kg/m2. Eighty-two participants were randomized into either the protein (intervention) or placebo (control) group. The exercise regime was the same for both groups. Out of 82 randomized participants, 58 completed the intervention. Blood tests were conducted for the amino acid OMICS measurement followed by dual-energy X-ray absorptiometry (DXA) for body composition and isokinetic dynamometry for muscle strength. A significant improvement was observed in the lean mass (kg) and appendicular muscle mass (AMM) adjusted for weight in the intervention group compared to the control group (p < 0.05). The muscle strength and contractile quality were comparable in the 2 groups. Plasma BCAA showed a significant negative association with body fat % (r = -0.43, p < 0.05 for the intervention group and r = -0.33, p = 0.07 for the control group) and a positive association with lean body mass % (r = 0.56, p < 0.01 in the intervention group vs r = 0.29, p = 0.10 in the control group) in the intervention group compared to control. In conclusion, this study highlighted the value of incorporating a lifestyle intervention including protein supplementation with simple exercises to optimize body composition in sedentary healthy individuals. This trial is registered with CTRI/2018/12/016777.

A machine‐learning approach to biomarker evaluation for AD precision medicine

AbstractBackgroundLate‐onset Alzheimer disease (AD) is a heterogeneous disease, as demonstrated by its wide range of genetic and environmental risk factors, as well as diversity of clinical manifestations. Cognizant of the futility of a one‐size‐fits‐all approach for heterogeneous diseases, precision medicine aims to treat each subtype appropriately, but methods for successful diagnosis of AD subtypes has been limited. Real‐valued biomarkers, such as omics measurements in plasma or CSF, frequently are evaluated based on fold change (FC) and/or the area under the receiver operating characteristic curve (AUC) metric. We previously demonstrated the inability for FC and AUC to capture signals for subtypes that comprise less than 50% of the diseased individuals due to inherently low true positive rates and proposed an alternative approach based on the bimodality of the data [Smith and Climer, https://doi.org/10.1101/2022.02.14.22270972]. This approach is based on statistical characteristics of the data, including skewness, cardinality, and kurtosis. k‐medians clustering, a machine learning approach to identify natural clusters in data, removes these statistical assumptions, but optimally solving k‐medians is generally NP‐hard, yielding computational intractability for datasets of interest. Lloyd’s approximation algorithm is commonly used in practice, however, its accuracy is highly sensitive to the random start state, thus is not suitable when accuracy is paramount.MethodOwing to the nature of the problem, we present an evaluation metric based on the optimal k‐medians objective with k = 2. Additionally, due to the single dimensionality of the data, we introduce a dynamic programming algorithm to improve computational performance.ResultAlthough optimally solving k‐medians is NP‐hard, our 2‐medians objective over 1‐dimensional data using dynamic programming has an average time complexity of O(n log n), and worst‐case complexity of O(n 2) where n is the number of data values. This objective has no reliance upon false positive rate, skewness, or kurtosis, making it robust for real‐valued data drawn from heterogeneous AD subtypes.ConclusionBiomarker data drawn from heterogeneous AD subtypes will have inherently low true positive rates, resulting with low FC and AUC values. Our machine‐learning based approach for evaluating potential biomarkers efficiently produces optimal clustering of the data and scales well to large datasets.

Multi-omic prediction of incident type 2 diabetes

Aims/hypothesisThe identification of people who are at high risk of developing type 2 diabetes is a key part of population-level prevention strategies. Previous studies have evaluated the predictive utility of omics measurements, such as metabolites, proteins or polygenic scores, but have considered these separately. The improvement that combined omics biomarkers can provide over and above current clinical standard models is unclear. The aim of this study was to test the predictive performance of genome, proteome, metabolome and clinical biomarkers when added to established clinical prediction models for type 2 diabetes.MethodsWe developed sparse interpretable prediction models in a prospective, nested type 2 diabetes case-cohort study (N=1105, incident type 2 diabetes cases=375) with 10,792 person-years of follow-up, selecting from 5759 features across the genome, proteome, metabolome and clinical biomarkers using least absolute shrinkage and selection operator (LASSO) regression. We compared the predictive performance of omics-derived predictors with a clinical model including the variables from the Cambridge Diabetes Risk Score and HbA1c.ResultsAmong single omics prediction models that did not include clinical risk factors, the top ten proteins alone achieved the highest performance (concordance index [C index]=0.82 [95% CI 0.75, 0.88]), suggesting the proteome as the most informative single omic layer in the absence of clinical information. However, the largest improvement in prediction of type 2 diabetes incidence over and above the clinical model was achieved by the top ten features across several omic layers (C index=0.87 [95% CI 0.82, 0.92], Δ C index=0.05, p=0.045). This improvement by the top ten omic features was also evident in individuals with HbA1c <42 mmol/mol (6.0%), the threshold for prediabetes (C index=0.84 [95% CI 0.77, 0.90], Δ C index=0.07, p=0.03), the group in whom prediction would be most useful since they are not targeted for preventative interventions by current clinical guidelines. In this subgroup, the type 2 diabetes polygenic risk score was the major contributor to the improvement in prediction, and achieved a comparable improvement in performance when added onto the clinical model alone (C index=0.83 [95% CI 0.75, 0.90], Δ C index=0.06, p=0.002). However, compared with those with prediabetes, individuals at high polygenic risk in this group had only around half the absolute risk for type 2 diabetes over a 20 year period.Conclusions/interpretationOmic approaches provided marginal improvements in prediction of incident type 2 diabetes. However, while a polygenic risk score does improve prediction in people with an HbA1c in the normoglycaemic range, the group in whom prediction would be most useful, even individuals with a high polygenic burden in that subgroup had a low absolute type 2 diabetes risk. This suggests a limited feasibility of implementing targeted population-based genetic screening for preventative interventions.Graphical

Sample Preparation Method for MALDI Mass Spectrometry Imaging of Fresh-Frozen Spines.

Technologies assessing the lipidomics, genomics, epigenomics, transcriptomics, and proteomics of tissue samples at single-cell resolution have deepened our understanding of physiology and pathophysiology at an unprecedented level of detail. However, the study of single-cell spatial metabolomics in undecalcified bones faces several significant challenges, such as the fragility of bone, which often requires decalcification or fixation leading to the degradation or removal of lipids and other molecules. As such, we describe a method for performing mass spectrometry imaging on undecalcified spine that is compatible with other spatial omics measurements. In brief, we use fresh-frozen rat spines and a system of carboxyl methylcellulose embedding, cryofilm, and polytetrafluoroethylene rollers to maintain tissue integrity while avoiding signal loss from variations in laser focus and artifacts from traditional tissue processing. This reveals various tissue types and lipidomic profiles of spinal regions at 10 μm spatial resolutions using matrix-assisted laser desorption/ionization mass spectrometry imaging. We expect this method to be adapted and applied to the analysis of the spinal cord, shedding light on the mechanistic aspects of cellular heterogeneity, development, and disease pathogenesis underlying different bone-related conditions and diseases. This study furthers the methodology for high spatial metabolomics of spines and adds to the collective efforts to achieve a holistic understanding of diseases via single-cell spatial multiomics.

Evaluation of input data modality choices on functional gene embeddings

Abstract Functional gene embeddings, numerical vectors capturing gene function, provide a promising way to integrate functional gene information into machine learning models. These embeddings are learnt by applying self-supervised machine-learning algorithms on various data types including quantitative omics measurements, protein–protein interaction networks and literature. However, downstream evaluations comparing alternative data modalities used to construct functional gene embeddings have been lacking. Here we benchmarked functional gene embeddings obtained from various data modalities for predicting disease-gene lists, cancer drivers, phenotype–gene associations and scores from genome-wide association studies. Off-the-shelf predictors trained on precomputed embeddings matched or outperformed dedicated state-of-the-art predictors, demonstrating their high utility. Embeddings based on literature and protein–protein interactions inferred from low-throughput experiments outperformed embeddings derived from genome-wide experimental data (transcriptomics, deletion screens and protein sequence) when predicting curated gene lists. In contrast, they did not perform better when predicting genome-wide association signals and were biased towards highly-studied genes. These results indicate that embeddings derived from literature and low-throughput experiments appear favourable in many existing benchmarks because they are biased towards well-studied genes and should therefore be considered with caution. Altogether, our study and precomputed embeddings will facilitate the development of machine-learning models in genetics and related fields.

Multi-omics characterization of NIST seafood reference materials and alternative matrix preparations.

The National Institute of Standards and Technology (NIST) has prepared four seafood reference materials (RMs) for use in food safety and nutrition studies: wild-caught and aquacultured salmon (RM 8256 and RM 8257) and wild-caught and aquacultured shrimp (RM 8258 and RM 8259). These materials were characterized using genetic, metabolomic (1H-NMR, nuclear magnetic resonance and LC-HRMS/MS, liquid chromatography high-resolution tandem mass spectrometry), lipidomic, and proteomic methods to explore their use as matrix-matched, multi-omic differential materials for method development towards identifying product source and/or as quality control in untargeted omics studies. The results from experimental replicates were reproducible for each reference material and analytical method, with the most abundant features reported. Additionally, differences between the materials could be detected, where wild-caught and aquacultured seafood could be distinguished using untargeted metabolite, lipid, and protein analyses. Further processing of the fresh-frozen RMs by freeze-drying revealed the freeze-dried seafoods could still be reliably discerned. These results demonstrate the usefulness of these reference materials as tools for omics instrument validation and measurement harmonization in seafood-related studies. Furthermore, their use as differential quality control (QC) materials, regardless of preparation method, may also provide a tool for laboratories to demonstrate proficiency at discriminating between products based on source/species.

Microbiability of milk composition and genetic control of microbiota effects in sheep

Recently, high-dimensional omics data are becoming available in larger quantities, and models have been developed that integrate them with genomics to understand in finer detail the relationship between genotype and phenotype, and thus improve the performance of genetic evaluations. Our objectives are to quantify the effect of the inclusion of microbiome data in the genetic evaluation for dairy traits in sheep, through the estimation of the heritability, microbiability, and how the microbiome effect on dairy traits decomposes into genetic and nongenetic parts. In this study we analyzed milk and rumen samples of 795 Lacaune dairy ewes. We included, as phenotype, dairy traits and milk fatty acids and proteins composition; as omics measurements, 16S rRNA rumen bacterial abundances; and as genotyping, 54K SNP chip for all ewes. Two nested genomic models were used: a first model to predict the individual contributions of the genetic and microbial abundances to phenotypes, and a second model to predict the additive genetic effect of the microbial community. In addition, microbiome-wide association studies for all dairy traits were applied using the 2,059 rumen bacterial abundances, and the genetic correlations between microbiome principal components and dairy traits were estimated. Results showed that in general the inclusion of both genetic and microbiome effect did not improve the fit of the model compared with the model with the genetic effect only. In addition, for all dairy traits the total heritability was equal to the direct heritability after fitting microbiota effects, due to a microbiability being almost zero for most dairy traits and heritability of the microbial community was very close to zero. Microbiome-wide association studies did not show operational taxonomic units with major effect for any of the dairy traits evaluated, and the genetic correlations between the first 5 principal components and dairy traits were low to moderate. So far, we can conclude that, using a substantial data set of 795 Lacaune dairy ewes, rumen bacterial abundances do not provide improved genetic evaluation for dairy traits in sheep.

Aligned deep neural network for integrative analysis with high-dimensional input

Objective:Deep neural network (DNN) techniques have demonstrated significant advantages over regression and some other techniques. In recent studies, DNN-based analysis has been conducted on data with high-dimensional input such as omics measurements. In such analysis, regularization, in particular penalization, has been applied to regularize estimation and distinguish relevant input variables from irrelevant ones. A unique challenge arises from the “lack of information” attributable to high dimensionality of input and limited size of training data. For many data/studies, there exist other data/studies that may be relevant and can potentially provide additional information to boost performance. Methods:In this study, we conduct integrative analysis of multiple independent datasets/studies, with the goal of borrowing information across each other and improving overall performance. Significantly different from regression-based integrative analysis (where alignment can be easily achieved based on covariates), alignment across multiple DNNs can be nontrivial. We develop ANNI, an Aligned DNN technique for Integrative analysis with high-dimensional input. Penalization is applied for regularized estimation, selection of important input variables, and, equally importantly, information borrowing across multiple DNNs. An effective computational algorithm is developed. Results:Extensive simulations demonstrate competitive performance of the proposed technique. The analysis of cancer omics data further establishes its practical utility.

Single-cell technologies for multimodal omics measurements

The recent surge in single-cell genomics, including the development of a wide range of experimental and computational approaches, has provided insights into the complex molecular networks of cells during development and in human diseases at unprecedented resolution. Single-cell transcriptome analysis has enabled high-resolution investigation of cellular heterogeneity in a wide range of cell populations ranging from early embryos to complex tissues—while posing the risk of only capturing a partial picture of the cells’ complex molecular networks. Single-cell multiomics technologies aim to bridge this gap by providing a more holistic view of the cell by simultaneously measuring multiple molecular types from the same cell and providing a more complete view of the interactions and combined functions of multiple regulatory layers at cell-type resolution. In this review, we briefly summarized the recent advances in multimodal single-cell technologies and discussed the challenges and opportunities of the field.

Multiomic signatures of body mass index identify heterogeneous health phenotypes and responses to a lifestyle intervention

Multiomic profiling can reveal population heterogeneity for both health and disease states. Obesity drives a myriad of metabolic perturbations and is a risk factor for multiple chronic diseases. Here we report an atlas of cross-sectional and longitudinal changes in 1,111 blood analytes associated with variation in body mass index (BMI), as well as multiomic associations with host polygenic risk scores and gut microbiome composition, from a cohort of 1,277 individuals enrolled in a wellness program (Arivale). Machine learning model predictions of BMI from blood multiomics captured heterogeneous phenotypic states of host metabolism and gut microbiome composition better than BMI, which was also validated in an external cohort (TwinsUK). Moreover, longitudinal analyses identified variable BMI trajectories for different omics measures in response to a healthy lifestyle intervention; metabolomics-inferred BMI decreased to a greater extent than actual BMI, whereas proteomics-inferred BMI exhibited greater resistance to change. Our analyses further identified blood analyte–analyte associations that were modified by metabolomics-inferred BMI and partially reversed in individuals with metabolic obesity during the intervention. Taken together, our findings provide a blood atlas of the molecular perturbations associated with changes in obesity status, serving as a resource to quantify metabolic health for predictive and preventive medicine.