Landmark Genes Research Articles

Characterizing diseases using genetic and clinical variables: A data analytics approach

AbstractPredictive analytics is crucial in precision medicine for personalized patient care. To aid in precision medicine, this study identifies a subset of genetic and clinical variables that can serve as predictors for classifying diseased tissues/disease types. To achieve this, experiments were performed on diseased tissues obtained from the L1000 dataset to assess differences in the functionality and predictive capabilities of genetic and clinical variables. In this study, the k‐means technique was used for clustering the diseased tissue types, and the multinomial logistic regression (MLR) technique was applied for classifying the diseased tissue types. Dimensionality reduction techniques including principal component analysis and Boruta are used extensively to reduce the dimensionality of genetic and clinical variables. The results showed that landmark genes performed slightly better in clustering diseased tissue types compared to any random set of 978 non‐landmark genes, and the difference is statistically significant. Furthermore, it was evident that both clinical and genetic variables were important in predicting the diseased tissue types. The top three clinical predictors for predicting diseased tissue types were identified as morphology, gender, and age of diagnosis. Additionally, this study explored the possibility of using the latent representations of the clusters of landmark and non‐landmark genes as predictors for an MLR classifier. The classification models built using MLR revealed that landmark genes can serve as a subset of genetic variables and/or as a proxy for clinical variables. This study concludes that combining predictive analytics with dimensionality reduction effectively identifies key predictors in precision medicine, enhancing diagnostic accuracy.

DOSE-L1000: unveiling the intricate landscape of compound-induced transcriptional changes.

The LINCS L1000 project has collected gene expression profiles for thousands of compounds across a wide array of concentrations, cell lines, and time points. However, conventional analysis methods often fall short in capturing the rich information encapsulated within the L1000 transcriptional dose-response data. We present DOSE-L1000, a database that unravels the potency and efficacy of compound-gene pairs and the intricate landscape of compound-induced transcriptional changes. Our study uses the fitting of over 140 million generalized additive models and robust linear models, spanning the complete spectrum of compounds and landmark genes within the LINCS L1000 database. This systematic approach provides quantitative insights into differential gene expression and the potency and efficacy of compound-gene pairs across diverse cellular contexts. Through examples, we showcase the application of DOSE-L1000 in tasks such as cell line and compound comparisons, along with clustering analyses and predictions of drug-target interactions. DOSE-L1000 fosters applications in drug discovery, accelerating the transition to omics-driven drug development. DOSE-L1000 is publicly available at https://doi.org/10.5281/zenodo.8286375.

Retinoic Acid Receptor Is a Novel Therapeutic Target for Postoperative Cognitive Dysfunction.

Postoperative cognitive dysfunction (POCD) is a clinical syndrome characterizing by cognitive impairments in the elderly after surgery. There is limited effective treatment available or clear pathological mechanisms known for this syndrome. In this study, a Connectivity Map (CMap) bioinformatics model of POCD was established by using differently expressed landmark genes in the serum samples of POCD and non-POCD patients from the only human transcriptome study. The predictability and reliability of this model were further supported by the positive CMap scores of known POCD inducers and the negative CMap scores of anti-POCD drug candidates. Most retinoic acid receptor (RAR) agonists were negatively associated with POCD in this CMap model, suggesting that RAR might be a novel target for POCD. Most importantly, acitretin, a clinically used RAR agonist, significantly inhibited surgery-induced cognitive impairments and prevented the reduction in RARα and RARα-target genes in the hippocampal regions of aged mice. The study denotes a reliable CMap bioinformatics model of POCD for future use and establishes that RAR is a novel therapeutic target for treating this clinical syndrome.

Gene expression prediction based on neighbour connection neural network utilizing gene interaction graphs.

Having observed that gene expressions have a correlation, the Library of Integrated Network-based Cell-Signature program selects 1000 landmark genes to predict the remaining gene expression value. Further works have improved the prediction result by using deep learning models. However, these models ignore the latent structure of genes, limiting the accuracy of the experimental results. We therefore propose a novel neural network named Neighbour Connection Neural Network(NCNN) to utilize the gene interaction graph information. Comparing to the popular GCN model, our model incorperates the graph information in a better manner. We validate our model under two different settings and show that our model promotes prediction accuracy comparing to the other models.

Transcriptomic points of departure calculated from human intestinal cells exposed to dietary nanoparticles.

Use of nanoparticles (NPs) in the food industry raises dietary health concerns. Assessing dietary NPs remains challenged due the vast number of products and the resource-intensive nature of toxicity testing. Advancements in high-throughput transcriptomics, coupled with benchmark dose (BMD) analysis are poised to modernize chemical safety assessments. The study objective was to derive transcriptomic point of departure (tPOD) values for common dietary NPs through dose-response analysis of 3'RNA-sequencing data. Two intestinal cell lines (Caco-2, HIEC-6) were exposed to 9 forms of Ag, SiO2, and TiO2, and expression of L1000 landmark genes was characterized. In Caco-2cells, tPODmode concentrations were 0.4-0.6, 21-32, and 17-59ppm for NPs of Ag, SiO2, and TiO2, respectively; in HIEC-6cells, the respective tPOD values were 6-7, 7-9, and 3-13ppm. Pathway BMDs across cases identified, for example, osteoclast and Th1/Th2 cell differentiation, and cell cycle, signaling, and senescence pathways. In all cases, the tPOD and pathway BMD values were lower than concentrations associated with cellular changes (e.g., generation of reactive oxygen species and proinflammatory cytokines, and cytotoxicity). These results demonstrate that transcriptomics dose-response analysis using in vitro models can help to increase understanding of a NP's mechanisms of action and derive quantitative information for dietary risk assessment.

Transforming L1000 profiles to RNA-seq-like profiles with deep learning

The L1000 technology, a cost-effective high-throughput transcriptomics technology, has been applied to profile a collection of human cell lines for their gene expression response to > 30,000 chemical and genetic perturbations. In total, there are currently over 3 million available L1000 profiles. Such a dataset is invaluable for the discovery of drug and target candidates and for inferring mechanisms of action for small molecules. The L1000 assay only measures the mRNA expression of 978 landmark genes while 11,350 additional genes are computationally reliably inferred. The lack of full genome coverage limits knowledge discovery for half of the human protein coding genes, and the potential for integration with other transcriptomics profiling data. Here we present a Deep Learning two-step model that transforms L1000 profiles to RNA-seq-like profiles. The input to the model are the measured 978 landmark genes while the output is a vector of 23,614 RNA-seq-like gene expression profiles. The model first transforms the landmark genes into RNA-seq-like 978 gene profiles using a modified CycleGAN model applied to unpaired data. The transformed 978 RNA-seq-like landmark genes are then extrapolated into the full genome space with a fully connected neural network model. The two-step model achieves 0.914 Pearson’s correlation coefficients and 1.167 root mean square errors when tested on a published paired L1000/RNA-seq dataset produced by the LINCS and GTEx programs. The processed RNA-seq-like profiles are made available for download, signature search, and gene centric reverse search with unique case studies.

Abstract 1196: Elucidating compound mechanism of action and polypharmacology with a large-scale perturbational profile compendium

Abstract Traditional approaches to elucidate small molecule mechanism of action (MoA) are usually based on affinity binding assays. These often fail to detect lower affinity binding targets and secondary effectors that are usually highly tissue context-dependent. As such, these can vary dramatically across different cancer types. To address this challenge and facilitate more quantitative approaches to targeted cancer therapy, a more comprehensive picture of drug MoA, including poly-pharmacology and toxicity effects mediated by the full range of its high-affinity (primary), as well as its context-specific lower-affinity (secondary) and downstream effector (tertiary) targets. For this purpose, we generated transcriptional genome-wide RNA-Seq profiles of cancer cell lines, following compounds perturbation with FDA approved and late-stage experimental oncology drugs using a fully automated and highly efficient PLATE-Seq technology1. Compared to prior approaches, this provides (a) genome-wide readouts, rather than readouts limited to a small set of landmark genes, (b) in cell lines that were specifically selected as the highest-fidelity models for published human cancer cohorts, using the OncoMatch algorithm2, and (c) for a drug repertoire that is clinically relevant. In addition, to avoid confounding effects from drug stress/death response pathway activation—a common issue in previous studies—drugs were titrated at their maximum sub-lethal concentration (48-hr IC20) using 10-point dose response curves in each cell line. The ~20,000 molecular profiles generated by these assays were analyzed using the VIPER algorithm3 to reproducibly assess the effect of each drug on the activity of ~6,500 regulatory and signaling proteins compared to vehicle control. The resulting PanACEA (Pancancer Activity-based Compound Efficacy Analysis), a database comprising drug perturbation profiles for 23 different cancer cell lines and > 700 oncology drugs, representing the largest resources of functionally annotated, genome-wide perturbational profiles for clinically relevant drugs. Systematic analysis identified critical drug mechanism of action and drugs capable of reproducibly targeting undruggable proteins, such as MYC or KRAS. We also leveraged graph theory-based network approach to generate a conserved drug functional network and drug functional networks in different cancer contexts, which pinpoint critical novel poly-pharmacology effects. Citation Format: Lucas Zhongming Hu, Eugene Douglass, Ron Realubit, Charles Karan, Mariano Alvarez, Andrea Califano. Elucidating compound mechanism of action and polypharmacology with a large-scale perturbational profile compendium [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1196.

A transcriptomic reevaluation of the accessory olfactory organ in Bichir (Polypterus senegalus)

Fish possess one olfactory organ called the olfactory epithelium (OE), by which various chemical substances are detected. On the other hand, tetrapods possess two independent olfactory organs called the main olfactory epithelium (MOE) and vomeronasal organ (VNO), each of which mainly detects general odorants and pheromones, respectively. Traditionally, the VNO, so-called concentrations of vomeronasal neurons, was believed to have originated in tetrapods. However, recent studies have identified a primordial VNO in lungfish, implying that the origin of the VNO was earlier than traditionally expected. In this study, we examined the presence/absence of the VNO in the olfactory organ of bichir (Polypterus senegalus), which is the most ancestral group of extant bony vertebrates. In particular, we conducted a transcriptomic evaluation of the accessory olfactory organ (AOO), which is anatomically separated from the main olfactory organ (MOO) in bichir. As a result, several landmark genes specific to the VNO and MOE in tetrapods were both expressed in the MOO and AOO, suggesting that these organs were not functionally distinct in terms of pheromone and odorant detection. Instead, differentially expressed gene (DEG) analysis showed that DEGs in AOO were enriched in genes for cilia movement, implying its additional and specific function in efficient water uptake into the nasal cavity other than chemosensing. This transcriptomic study provides novel insight into the long-standing question of AOO function in bichir and suggests that VNO originated in the lineage of lobe-finned fish during vertebrate evolution.

SWnet: a deep learning model for drug response prediction from cancer genomic signatures and compound chemical structures

BackgroundOne of the major challenges in precision medicine is accurate prediction of individual patient’s response to drugs. A great number of computational methods have been developed to predict compounds activity using genomic profiles or chemical structures, but more exploration is yet to be done to combine genetic mutation, gene expression, and cheminformatics in one machine learning model.ResultsWe presented here a novel deep-learning model that integrates gene expression, genetic mutation, and chemical structure of compounds in a multi-task convolutional architecture. We applied our model to the Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Cell Line Encyclopedia (CCLE) datasets. We selected relevant cancer-related genes based on oncology genetics database and L1000 landmark genes, and used their expression and mutations as genomic features in model training. We obtain the cheminformatics features for compounds from PubChem or ChEMBL. Our finding is that combining gene expression, genetic mutation, and cheminformatics features greatly enhances the predictive performance.ConclusionWe implemented an extended Graph Neural Network for molecular graphs and Convolutional Neural Network for gene features. With the employment of multi-tasking and self-attention functions to monitor the similarity between compounds, our model outperforms recently published methods using the same training and testing datasets.

Enhancing performance of gene expression value prediction with cluster-based regression.

The inherent correlations among gene expressions have received attention. Recently, it was reported that a set of approximately 1000 landmark genes can be utilized for prediction of expression of other genes (target genes). The objective of this study is to predict expression values of target genes based on expression values of landmark genes. A cluster-based regression method is proposed. In the proposed method, clusters are obtained from a set of training instances of a gene and an estimator is obtained per cluster. A test instance is assigned to one of clusters then a regression model corresponding to the cluster predicts expression value. Performance of the proposed method is measured on the GEO (Gene Expression Omnibus) expression data and the GTEx (Genotype-Tissue Expression) expression data. In terms of mean absolute error averaged across target genes, the proposed method significantly outperforms previous approaches in the case of the GEO expression data. The experimental results report that the combination of clustering and regression can outperform the state-of-the art methods such as generative adversarial networks and a gradient boosting based method.

Single-cell BCR and transcriptome analysis after influenza infection reveals spatiotemporal dynamics of antigen-specific B cells.

SummaryB cell responses are critical for antiviral immunity. However, a comprehensive picture of antigen-specific B cell differentiation, clonal proliferation, and dynamics in different organs after infection is lacking. Here, by combining single-cell RNA and B cell receptor (BCR) sequencing of antigen-specific cells in lymph nodes, spleen, and lungs after influenza infection in mice, we identify several germinal center (GC) B cell subpopulations and organ-specific differences that persist over the course of the response. We discover transcriptional differences between memory cells in lungs and lymphoid organs and organ-restricted clonal expansion. Remarkably, we find significant clonal overlap between GC-derived memory and plasma cells. By combining BCR-mutational analyses with monoclonal antibody (mAb) expression and affinity measurements, we find that memory B cells are highly diverse and can be selected from both low- and high-affinity precursors. By linking antigen recognition with transcriptional programming, clonal proliferation, and differentiation, these finding provide important advances in our understanding of antiviral immunity.

Clump sequencing exposes the spatial expression programs of intestinal secretory cells

Single-cell RNA sequencing combined with spatial information on landmark genes enables reconstruction of spatially-resolved tissue cell atlases. However, such approaches are challenging for rare cell types, since their mRNA contents are diluted in the spatial transcriptomics bulk measurements used for landmark gene detection. In the small intestine, enterocytes, the most common cell type, exhibit zonated expression programs along the crypt-villus axis, but zonation patterns of rare cell types such as goblet and tuft cells remain uncharacterized. Here, we present ClumpSeq, an approach for sequencing small clumps of attached cells. By inferring the crypt-villus location of each clump from enterocyte landmark genes, we establish spatial atlases for all epithelial cell types in the small intestine. We identify elevated expression of immune-modulatory genes in villus tip goblet and tuft cells and heterogeneous migration patterns of enteroendocrine cells. ClumpSeq can be applied for reconstructing spatial atlases of rare cell types in other tissues and tumors.

Fractal Autoencoders for Feature Selection

Feature selection reduces the dimensionality of data by identifying a subset of the most informative features. In this paper, we propose an innovative framework for unsupervised feature selection, called fractal autoencoders (FAE). It trains a neural network to pinpoint informative features for global exploring of representability and for local excavating of diversity. Architecturally, FAE extends autoencoders by adding a one-to-one scoring layer and a small sub-neural network for feature selection in an unsupervised fashion. With such a concise architecture, FAE achieves state-of-the-art performances; extensive experimental results on fourteen datasets, including very high-dimensional data, have demonstrated the superiority of FAE over existing contemporary methods for unsupervised feature selection. In particular, FAE exhibits substantial advantages on gene expression data exploration, reducing measurement cost by about 15% over the widely used L1000 landmark genes. Further, we show that the FAE framework is easily extensible with an application.

Deep Large-Scale Multitask Learning Network for Gene Expression Inference.

Gene expression profiling makes it possible to conduct many biological studies in a variety of fields due to its thorough characterization of cellular states under various experimental conditions. Despite recent advances in high-throughput technology, profiling an entire set of genomes is still difficult and expensive. Due to the high correlation between expression patterns of different genes, the aforementioned problem can be solved with a cost-effective approach that collects only a small subset of genes, called landmark genes, representing the entire set of genes, and infer the remaining genes, called target genes, using a computational model. There are several shallow and deep regression models in literature to estimate the expressions of target genes from the landmark genes. However, the shallow mostly have limited capacity in learning the nonlinear and complex gene expression data and are prone to underfitting, and the deep models generally do not take advantage of correlation among target genes in the learning process and suffer from overfitting. Considering the gene expression inference as a multitask learning problem, we propose a new deep multitask learning algorithm to tackle these issues. Our learning framework automatically learns the correlation between target genes and uses this knowledge to improve its generalization. Specifically, we utilize a subnetwork with low-dimensional latent variables to discover the relationships between target genes and enforce a seamless and easy to implement regularization to our deep regression model. Unlike the existing multitask learning methods that can only deal with dozens or hundreds of tasks, our algorithm is able to efficiently learn the relationships between ∼10,000 target genes and, thus, is scalable to a large number of tasks. Our proposed method outperforms the shallow and deep regression models for gene expression inference and alternative multitask learning algorithms on two large-scale datasets regardless of the network architecture.

Spatial mapping of single cells in the Drosophila embryo from transcriptomic data based on topological consistency.

The advancement in single-cell RNA sequencing technologies allow us to obtain transcriptome at single cell resolution. However, the original spatial context of cells, a crucial knowledge for understanding cellular and tissue-level functions, is often lost during sequencing. To address this issue, the DREAM Single Cell Transcriptomics Challenge launched a community-wide effort to seek computational solutions for spatial mapping of single cells in tissues using single-cell RNAseq (scRNA-seq) data and a reference atlas obtained from in situ hybridization data. As a top-performing team in this competition, we approach this problem in three steps. The first step involves identifying a set of most informative genes based on the consistency between gene expression similarity and cell proximity. For this step, we propose two different approaches, i.e., an unsupervised approach that does not utilize the gold standard location of the cells provided by the challenge organizers, and a supervised approach that relies on the gold standard locations. In the second step, a Particle Swarm Optimization algorithm is used to optimize the weights of different genes in order to maximize matches between the predicted locations and the gold standard locations. Finally, the information embedded in the cell topology is used to improve the predicted cell-location scores by weighted averaging of scores from neighboring locations. Evaluation results based on DREAM scores show that our method accurately predicts the location of single cells, and the predictions lead to successful recovery of the spatial expression patterns for most of landmark genes. In addition, investigating the selected genes demonstrates that most predictive genes are cluster specific, and stable across our supervised and unsupervised gene selection frameworks. Overall, the promising results obtained by our methods in DREAM challenge demonstrated that topological consistency is a useful concept in identifying marker genes and constructing predictive models for spatial mapping of single cells.

Mining influential genes based on deep learning

BackgroundCurrently, large-scale gene expression profiling has been successfully applied to the discovery of functional connections among diseases, genetic perturbation, and drug action. To address the cost of an ever-expanding gene expression profile, a new, low-cost, high-throughput reduced representation expression profiling method called L1000 was proposed, with which one million profiles were produced. Although a set of ~ 1000 carefully chosen landmark genes that can capture ~ 80% of information from the whole genome has been identified for use in L1000, the robustness of using these landmark genes to infer target genes is not satisfactory. Therefore, more efficient computational methods are still needed to deep mine the influential genes in the genome.ResultsHere, we propose a computational framework based on deep learning to mine a subset of genes that can cover more genomic information. Specifically, an AutoEncoder framework is first constructed to learn the non-linear relationship between genes, and then DeepLIFT is applied to calculate gene importance scores. Using this data-driven approach, we have re-obtained a landmark gene set. The result shows that our landmark genes can predict target genes more accurately and robustly than that of L1000 based on two metrics [mean absolute error (MAE) and Pearson correlation coefficient (PCC)]. This reveals that the landmark genes detected by our method contain more genomic information.ConclusionsWe believe that our proposed framework is very suitable for the analysis of biological big data to reveal the mysteries of life. Furthermore, the landmark genes inferred from this study can be used for the explosive amplification of gene expression profiles to facilitate research into functional connections.

On transformative adaptive activation functions in neural networks for gene expression inference.

Gene expression profiling was made more cost-effective by the NIH LINCS program that profiles only ∼1, 000 selected landmark genes and uses them to reconstruct the whole profile. The D-GEX method employs neural networks to infer the entire profile. However, the original D-GEX can be significantly improved. We propose a novel transformative adaptive activation function that improves the gene expression inference even further and which generalizes several existing adaptive activation functions. Our improved neural network achieves an average mean absolute error of 0.1340, which is a significant improvement over our reimplementation of the original D-GEX, which achieves an average mean absolute error of 0.1637. The proposed transformative adaptive function enables a significantly more accurate reconstruction of the full gene expression profiles with only a small increase in the complexity of the model and its training procedure compared to other methods.

An ensemble learning approach for modeling the systems biology of drug-induced injury

BackgroundDrug-induced liver injury (DILI) is an adverse reaction caused by the intake of drugs of common use that produces liver damage. The impact of DILI is estimated to affect around 20 in 100,000 inhabitants worldwide each year. Despite being one of the main causes of liver failure, the pathophysiology and mechanisms of DILI are poorly understood. In the present study, we developed an ensemble learning approach based on different features (CMap gene expression, chemical structures, drug targets) to predict drugs that might cause DILI and gain a better understanding of the mechanisms linked to the adverse reaction.ResultsWe searched for gene signatures in CMap gene expression data by using two approaches: phenotype-gene associations data from DisGeNET, and a non-parametric test comparing gene expression of DILI-Concern and No-DILI-Concern drugs (as per DILIrank definitions). The average accuracy of the classifiers in both approaches was 69%. We used chemical structures as features, obtaining an accuracy of 65%. The combination of both types of features produced an accuracy around 63%, but improved the independent hold-out test up to 67%. The use of drug-target associations as feature obtained the best accuracy (70%) in the independent hold-out test.ConclusionsWhen using CMap gene expression data, searching for a specific gene signature among the landmark genes improves the quality of the classifiers, but it is still limited by the intrinsic noise of the dataset. When using chemical structures as a feature, the structural diversity of the known DILI-causing drugs hampers the prediction, which is a similar problem as for the use of gene expression information. The combination of both features did not improve the quality of the classifiers but increased the robustness as shown on independent hold-out tests. The use of drug-target associations as feature improved the prediction, specially the specificity, and the results were comparable to previous research studies.

Tissue Classification Using Landmark and Non-Landmark Gene Sets for Feature Selection

The L1000 dataset, containing gene microarray data from 978 landmark genes has been previously shown to accurately predict expression of ~81% of the remaining 21,290 target genes. Microarray data was utilized to characterize groups of tissue types within the L1000 dataset to assess whether 978 landmark genes, compared to non-landmark genes, would better differentiate samples into clusters containing distinct tissue types. Landmark genes better differentiated k-means clusters, compared to non-landmark genes. These results suggest that landmark genes better characterize heterogeneous samples in their comprehensive genetic profile. Our previous studies showed that categorical separation of samples based on clinical or biological groups generally improves when studying heterogeneous sample types when using landmark genes as features, compared to non-landmark genes. However, the present work indicates that non-landmark genes may also be utilized to separate samples in clustering when there is a large sample size present for training k-means clustering models. In contrast, when studying a small sample size of the same set of heterogenous samples, landmark genes as features improve clustering. This study has implications for assessing various tissue types as landmark genes may be directly measured to predict categorical sample qualities as well as expression of remaining target genes.

Comparative Study of Disease Classification Using Multiple Machine Learning Models Based on Landmark and Non-Landmark Gene Expression Data

This study compares disease classification based on landmark and non-landmark gene expression data, and clinical variable using multiple machine-learning models. The influence of the number of principal components and the genes were also investigated. The results indicate that the ANN model has the best accuracy for disease type prediction among all the models, model using 95 principal components has better accuracy than that of 25 principal components, and the greater number of genes used, the higher the prediction accuracy. Models using landmark genes demonstrated better accuracy than the models using non-landmark genes especially with 95 PCs across all the models except for the decision trees. The optimal model was one that uses landmark genes with 95 PCs as features for an ANN classifier. The AUC measures obtained on the test set were 0.98,0.98,1 and 0.96 for Autoimmune, Bacteremia, Cancer and Healthy classes respectively, and the accuracy for the respective classes were 97.56%, 95.65%, 95.65%, and 58.82%. The ANN model demonstrated a good capability of distinguishing between the true positives and the false positives, and it resulted in high prediction accuracy for the 3 disease classes ( Autoimmune, Bacteremia, Cancer ), but it misclassified some instances from the Healthy class to the Autoimmune and Bacteremia class, likely due to a wide range of gene expression level for the Healthy class.