Biology subjects

Szalai, B.

Publications and source records attributed to Szalai, B..

15 recordsLinked to original sources

bioRxiv2024

RIDDEN: Data-driven inference of receptor activity from transcriptomic data

Intracellular signaling initiated from ligand bound receptors plays a fundamental role in both physiological regulation and development of disease states, making receptors one of the most frequent drug targets. Systems level analysis of receptor activity can help to identify cell and disease type specific receptor activity alterations. While recently several computational methods have been developed to analyse ligand -receptor interactions based on transcriptomics data, none of them focuses directly on the receptor side of these interactions. Also, most of the methods use directly the expression of ligands and receptors to infer active interaction, while co-expression of genes does not necessarily indicate functional interactions or activated state. To address these problems, we developed RIDDEN (Receptor actIvity Data Driven inferENce), a computational tool, which predicts receptor activities from the receptor-regulated gene expression profiles, and not from the expressions of ligand and receptor genes. We collected 14463 perturbation gene expression profiles for 229 different receptors. Using these data, we trained the RIDDEN model, which can effectively predict receptor activity for new bulk and single-cell transcriptomics datasets. We validated RIDDENs performance on independent in vitro and in vivo receptor perturbation data, showing that RIDDENs model weights correspond to known regulatory interactions between receptors and transcription factors, and that predicted receptor activities correlate with receptor and ligand expressions in in vivo datasets. We also show that RIDDEN can be used to identify mechanistic biomarkers in an immune checkpoint blockade-treated cancer patient cohort. RIDDEN, the largest transcriptomics-based receptor activity inference model, can be used to identify cell populations with altered receptor activity and, in turn, foster the study of cell-cell communication using transcriptomics data. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=157 SRC="FIGDIR/small/626558v1_ufig1.gif" ALT="Figure 1"> View larger version (31K): org.highwire.dtl.DTLVardef@155d87dorg.highwire.dtl.DTLVardef@f80ed2org.highwire.dtl.DTLVardef@195140eorg.highwire.dtl.DTLVardef@57ccfa_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2024

Benchmarking a foundational cell model for post-perturbation RNAseq prediction

Accurately predicting cellular responses to perturbations is essential for understanding cell behaviour in both healthy and diseased states. While perturbation data is ideal for building such predictive models, it is considerably sparser than baseline (non-perturbed) cellular data. To address this limitation, several foundational cell models have been developed using large-scale single-cell gene expression data. These models are fine-tuned after pre-training for specific tasks, such as predicting post-perturbation gene expression profiles, and are considered state-of-the-art for these problems. However, proper benchmarking of these models remains an unsolved challenge. In this study, we benchmarked a recently published foundational model, scGPT, against baseline models. Surprisingly, we found that even the simplest baseline model - taking the mean of training examples - outperformed scGPT. Furthermore, machine learning models that incorporate biologically meaningful features outperformed scGPT by a large margin. Additionally, we identified that the current Perturb-Seq benchmark datasets exhibit low perturbation-specific variance, making them suboptimal for evaluating such models. Our results highlight important limitations in current benchmarking approaches and provide insights into more effectively evaluating post-perturbation gene expression prediction models.

systems biology↗

bioRxiv2023

The EFFECT benchmark suite: measuring cancer sensitivity prediction performance - without the bias

1.Creating computational biology models applicable to industry is much more difficult than it appears. There is a major gap between a model that looks good on paper and a model that performs well in the drug discovery process. We are trying to shrink this gap by introducing the Evaluation Framework For predicting Efficiency of Cancer Treatment (EFFECT) benchmark suite based on the DepMap and GDSC data sets to facilitate the creation of well-applicable machine learning models capable of predicting gene essentiality and/or drug sensitivity on in vitro cancer cell lines. We show that standard evaluation metrics like Pearson correlation are misleading due to inherent biases in the data. Thus, to assess the performance of models properly, we propose the use of cell line/perturbation exclusive data splits, perturbation-wise evaluation, and the application of our Bias Detector framework, which can identify model predictions not explicable by data bias alone. Testing the EFFECT suite on a few popular machine learning (ML) models showed that while library-standard non-linear models have measurable performance in splits representing precision medicine and target identification tasks, the actual corrected correlations are rather low, showing that even simple knock-out (KO)/drug sensitivity prediction is a yet unsolved task. For this reason, we aim our proposed framework to be a unified test and evaluation pipeline for ML models predicting cancer sensitivity data, facilitating unbiased benchmarking to support teams to improve on the state of the art.

systems biology↗

bioRxiv2023

Receptor endocytosis orchestrates the spatiotemporal bias of β-arrestin signaling

The varying efficacy of biased and balanced agonists is generally explained by the stabilization of different active receptor conformations. In this study, systematic profiling of transducer activation of AT1 angiotensin receptor agonists revealed that the extent and kinetics of {beta}-arrestin binding exhibit substantial ligand-dependent differences, which however completely disappear upon the inhibition of receptor internalization. Even weak partial agonists for the {beta}- arrestin pathway acted as full or near full agonists, if receptor endocytosis was prevented, indicating that receptor conformation is not an exclusive determinant of {beta}-arrestin recruitment. The ligand-dependent variance in {beta}-arrestin translocation at endosomes was much larger than it was at the plasma membrane, showing that ligand efficacy in the {beta}-arrestin pathway is spatiotemporally determined. Experimental investigations and mathematical modeling demonstrated how multiple factors concurrently shape the effects of agonists on endosomal receptor-{beta}-arrestin binding and thus determine the extent of bias. Among others, ligand dissociation rate and G protein activity have particularly strong impact on receptor-{beta}-arrestin interaction, and their effects are integrated at endosomes. Our results highlight that endocytosis forms a key spatiotemporal platform for biased GPCR signaling and can aid the development of more efficacious functionally-selective compounds. One Sentence summaryAgonist-specific differences in {beta}-arrestin recruitment are mainly determined by the ligand dissociation rate and G protein activation at the endosomes.

cell biology↗

bioRxiv2021

Combi-Seq: Multiplexed transcriptome-based profiling of drug combinations using 1deterministic barcoding in single-cell droplets

Anti-cancer therapies often exhibit only short-term effects. Tumors typically develop drug resistance causing relapses that might be tackled with drug combinations. Identification of the right combination is challenging and would benefit from high-content, high-throughput combinatorial screens directly on patient biopsies. However, such screens require a large amount of material, normally not available from patients. To address these challenges, we developed a scalable microfluidic workflow to screen hundreds of drug combinations in picoliter-size droplets using transcriptome changes as a readout for drug effects. We devised a deterministic combinatorial DNA barcoding approach to encode treatment conditions, enabling the gene expression-based readout of drug effects in a highly multiplexed fashion. We applied our method to screen the effect of 420 drug combinations on the transcriptome of K562 cells using only [~]250 single cell droplets per condition, to successfully predict synergistic and antagonistic drug pairs, as well as their pathway activities.

systems biology↗

bioRxiv2021

Computational drug repurposing against SARS-CoV-2 reveals plasma membrane cholesterol depletion as key factor of antiviral drug activity

Comparing SARS-CoV-2 infection-induced gene expression signatures to drug treatment-induced gene expression signatures is a promising bioinformatic tool to repurpose existing drugs against SARS-CoV-2. The general hypothesis of signature based drug repurposing is that drugs with inverse similarity to a disease signature can reverse disease phenotype and thus be effective against it. However, in the case of viral infection diseases, like SARS-CoV-2, infected cells also activate adaptive, antiviral pathways, so that the relationship between effective drug and disease signature can be more ambiguous. To address this question, we analysed gene expression data from in vitro SARS-CoV-2 infected cell lines, and gene expression signatures of drugs showing anti-SARS-CoV-2 activity. Our extensive functional genomic analysis showed that both infection and treatment with in vitro effective drugs leads to activation of antiviral pathways like NFkB and JAK-STAT. Based on the similarity - and not inverse similarity - between drug and infection-induced gene expression signatures, we were able to predict the in vitro antiviral activity of drugs. We also identified SREBF1/2, key regulators of lipid metabolising enzymes, as the most activated transcription factors by several in vitro effective antiviral drugs. Using a fluorescently labeled cholesterol sensor, we showed that these drugs decrease the cholesterol levels of plasma-membrane. Supplementing drug-treated cells with cholesterol reversed the in vitro antiviral effect, suggesting the depleting plasma-membrane cholesterol plays a key role in virus inhibitory mechanism. Our results can help to more effectively repurpose approved drugs against SARS-CoV-2, and also highlights key mechanisms behind their antiviral effect. O_FIG O_LINKSMALLFIG WIDTH=171 HEIGHT=200 SRC="FIGDIR/small/459786v1_ufig1.gif" ALT="Figure 1"> View larger version (37K): org.highwire.dtl.DTLVardef@7cd823org.highwire.dtl.DTLVardef@51f699org.highwire.dtl.DTLVardef@114c555org.highwire.dtl.DTLVardef@a774ee_HPS_FORMAT_FIGEXP M_FIG C_FIG

systems biology↗

bioRxiv2021

Patient-specific Boolean models of signaling networks guide personalized treatments

Prostate cancer is the second most occurring cancer in men worldwide. To better understand the mechanisms of tumorigenesis and possible treatment responses, we developed a mathematical model of prostate cancer which considers the major signalling pathways known to be deregulated. We personalised this Boolean model to molecular data to reflect the heterogeneity and specific response to perturbations of cancer patients. 488 prostate samples were used to build patient-specific models and compared to available clinical data. Additionally, eight prostate cell-line-specific models were built to validate our approach with dose-response data of several drugs. The effects of single and combined drugs were tested in these models under different growth conditions. We identified 15 actionable points of interventions in one cell-line-specific model whose inactivation hinders tumorigenesis. To validate these results, we tested nine small molecule inhibitors of five of those putative targets and found a dose-dependent effect on four of them, notably those targeting HSP90 and PI3K. These results highlight the predictive power of our personalized Boolean models and illustrate how they can be used for precision oncology.

systems biology↗

bioRxiv2021

Comparison of Resources and Methods to infer Cell-Cell Communication from Single-cell RNA Data

The growing availability of single-cell data has sparked an increased interest in the inference of cell-cell communication from this data. Many tools have been developed for this purpose. Each of them consists of a resource of intercellular interactions prior knowledge and a method to predict potential cell-cell communication events. Yet the impact of the choice of resource and method on the resulting predictions is largely unknown. To shed light on this, we created a framework, available at https://github.com/saezlab/ligrec_decoupler, to facilitate a comparative assessment of methods for inferring cell-cell communication from single cell transcriptomics data and then compared 15 resources and 6 methods. We found few unique interactions and a varying degree of overlap among the resources, and observed uneven coverage in terms of pathways and biological categories. We analysed a colorectal cancer single cell RNA-Seq dataset using all possible combinations of methods and resources. We found major differences among the highest ranked intercellular interactions inferred by each method even when using the same resources. The varying predictions lead to fundamentally different biological interpretations, highlighting the need to benchmark resources and methods. FindingsO_LIBuilt a framework to systematically combine 15 resources and 6 methods to estimate cell-cell communication from single-cell RNA data C_LIO_LICell-cell communication resources are often built from the same original databases and very few interactions are unique to a single resource. Yet overlap varies among resources and certain biological terms are unevenly represented C_LIO_LIDifferent methods and resources provided notably different results C_LIO_LIThe observed disagreement among the methods could have a considerable impact on the interpretation of results C_LI

bioinformatics↗

bioRxiv2020

A Community Challenge for Pancancer Drug Mechanism of Action Inference from Perturbational Profile Data

The Columbia Cancer Target Discovery and Development (CTD2) Center has developed PANACEA (PANcancer Analysis of Chemical Entity Activity), a collection of dose-response curves and perturbational profiles for 400 clinical oncology drugs in cell lines selected to optimally represent 19 cancer subtypes. This resource, developed to study tumor-specific drug mechanism of action, was instrumental in hosting a DREAM Challenge to assess computational models for de novo drug polypharmacology prediction. Dose-response and perturbational profiles for 32 kinase inhibitors were provided to 21 participating teams, who did not know the identity or nature of the compounds, and they were asked to predict high-affinity binding among ~1,300 possible protein targets. Best performing methods leveraged both gene expression profile similarity analysis, and deep-learning methodologies trained on individual datasets. This study lays the foundation for future integrative analyses of pharmacogenomic data, reconciliation of polypharmacology effects in different tumor contexts, and insights into network-based assessment of context-specific drug mechanism of action.

cancer biology↗

bioRxiv2020

Why do pathway methods work better than they should?

Different pathway analysis methods are frequently applied to cancer gene expression data to identify dysregulated pathways. In most cases these methods infer pathway activity changes based on the gene expression of pathway members. However, pathways are constituted by signaling proteins, and their activity - not their abundance - defines the activity of the pathway; the association between gene expression and protein activity is in turn limited and not well characterised. Other methods infer pathway activity from the expression of the genes whose transcription is regulated by the pathway of interest, which seems a more adequate proxy of activity. Despite these potential limitations, membership based pathway methods are frequently used and often provide statistically significant results. Here, we submit that pathway based methods are not effective because of the correlation between the gene expression of pathway members and the activity of the pathway, but because pathway member gene sets overlap with the genes regulated by transcription factors (regulons). This implies that pathway methods do not inform about the activity of the pathway of interest, but instead the downstream effects of changes in the activities of transcription factors. To support our hypothesis, we show that the higher the overlap to transcription factor regulons, the higher the information value of pathway gene sets. Furthermore, removing these overlapping genes reduces the information content of pathway gene sets, but not vice versa. Our results suggest that results of classical pathway analysis methods should be interpreted with caution, and instead methods using pathway regulated genes for activity inference should be prioritised. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=129 SRC="FIGDIR/small/228296v1_ufig1.gif" ALT="Figure 1"> View larger version (25K): org.highwire.dtl.DTLVardef@f955baorg.highwire.dtl.DTLVardef@16b2f0aorg.highwire.dtl.DTLVardef@1dd6b7dorg.highwire.dtl.DTLVardef@1bbfe59_HPS_FORMAT_FIGEXP M_FIG C_FIG

systems biology↗

bioRxiv2019

Robustness and applicability of functional genomics tools on scRNA-seq data

Many tools have been developed to extract functional and mechanistic insight from bulk transcriptome profiling data. With the advent of single-cell RNA sequencing (scRNA-seq), it is in principle possible to do such an analysis for single cells. However, scRNA-seq data has characteristics such as drop-out events, low library sizes and a comparatively large number of samples/cells. It is thus not clear if functional genomics tools established for bulk sequencing can be applied to scRNA-seq in a meaningful way. To address this question, we performed benchmark studies on in silico and in vitro single-cell RNA-seq data. We included the bulk-RNA tools PROGENy, GO enrichment and DoRothEA that estimate pathway and transcription factor (TF) activities, respectively, and compared them against the tools AUCell and metaVIPER, designed for scRNA-seq. For the in silico study we simulated single cells from TF/pathway perturbation bulk RNA-seq experiments. Our simulation strategy guarantees that the information of the original perturbation is preserved while resembling the characteristics of scRNA-seq data. We complemented the in silico data with in vitro scRNA-seq data upon CRISPR-mediated knock-out. Our benchmarks on both the simulated and real data revealed comparable performance to the original bulk data. Additionally, we showed that the TF and pathway activities preserve cell-type specific variability by analysing a mixture sample sequenced with 13 scRNA-seq different protocols. Our analyses suggest that bulk functional genomics tools can be applied to scRNA-seq data, outperforming dedicated single cell tools. Furthermore we provide a benchmark for further methods development by the community.

bioinformatics↗

bioRxiv2019

Transfer of regulatory knowledge from human to mouse for functional genomic analysis

Transcriptome profiling followed by differential gene expression analysis often leads to unclear lists of genes which are hard to analyse and interpret. Functional genomic tools are powerful approaches for downstream analysis, as they summarize the large and noisy gene expression space in a smaller number of biological meaningful features. In particular, methods that estimate the activity of processes by mapping transcripts level to process members are popular. However, footprints of either a pathway or transcription factor (TF) on gene expression show superior performance over mapping-based gene sets. These footprints are largely developed for human and their usability in the broadly-used model organism Mus musculus is uncertain. Evolutionary conservation of the gene regulatory system suggests that footprints of human pathways and TFs can functionally characterize mice data. In this paper we analyze this hypothesis. We perform a comprehensive benchmark study exploiting two state-of-the-art footprint methods, DoRothEA and an extended version of PROGENy. These methods infer TF and pathway activity, respectively. Our results show that both can recover mouse perturbations, confirming our hypothesis that footprints are conserved between mice and humans. Subsequently, we illustrate the usability of PROGENy and DoRothEA by recovering pathway/TF-disease associations from newly generated disease sets. Additionally, we provide pathway and TF activity scores for a large collection of human and mouse perturbation and disease experiments (2,374). We believe that this resource, available for interactive exploration and download (https://saezlab.shinyapps.io/footprint_scores/), can have broad applications including the study of diseases and therapeutics.

systems biology↗

bioRxiv2018

Signatures of cell death and proliferation in perturbation transcriptomics data - from confounding factor to effective prediction

Transcriptomics perturbation signatures are valuable data sources for functional genomic studies. They can be effectively used to identify mechanism of action for new compounds and to infer functional activity of different cellular processes. Linking perturbation signatures to phenotypic studies opens up the possibility to model selected cellular phenotypes from gene expression data and to predict drugs interfering with the phenotype. At the same time, close association of transcriptomics changes with phenotypes can potentially mask the compound specific signatures. By linking perturbation transcriptomics data from the LINCS-L1000 project with cell viability phenotypic information upon genetic (from Achilles project) and chemical (from CTRP screen) perturbations for more than 90,000 signature - cell viability pairs, we show here that a cell death signature is a major factor behind perturbation signatures. We use this relationship to effectively predict cell viability from transcriptomics signatures, and identify and experimentally validate compounds that induce either cell death or proliferation. We also show that cellular toxicity can lead to an unexpected similarity of toxic compound signatures confounding the mechanism of action discovery. Consensus compound signatures predict cell-specific anti-cancer drug sensitivity, even if the drug signature is not measured in the same cell line. These signatures outperform conventional drug-specific features like nominal target and chemical fingerprints. Our results can help removing confounding factors of large scale transcriptomics perturbation screens and show that expression signatures boost prediction of drug sensitivity.

systems biology↗

bioRxiv2017

Community assessment of cancer drug combination screens identifies strategies for synergy prediction

The effectiveness of most cancer targeted therapies is short lived since tumors evolve and develop resistance. Combinations of drugs offer the potential to overcome resistance, however the number of possible combinations is vast necessitating data-driven approaches to find optimal treatments tailored to a patients tumor. AstraZeneca carried out 11,576 experiments on 910 drug combinations across 85 cancer cell lines, recapitulating in vivo response profiles. These data, the largest openly available screen, were hosted by DREAM alongside deep molecular characterization from the Sanger Institute for a Challenge to computationally predict synergistic drug pairs and associated biomarkers. 160 teams participated to provide the most comprehensive methodological development and subsequent benchmarking to date. Winning methods incorporated prior knowledge of putative drug target interactions. For >60% of drug combinations synergy was reproducibly predicted with an accuracy matching biological replicate experiments, however 20% of drug combinations were poorly predicted by all methods. Genomic rationale for synergy predictions were identified, including antagonism unique to combined PIK3CB/D inhibition with the ADAM17 inhibitor where synergy is seen with other PI3K pathway inhibitors. All data, methods and code are freely available as a resource to the community.

bioinformatics↗

bioRxiv2016

Reverse-engineering human olfactory perception from chemical features of odor molecules

Despite 25 years of progress in understanding the molecular mechanisms of olfaction, it is still not possible to predict whether a given molecule will have a perceived odor, or what olfactory percept it will produce. To address this stimulus-percept problem for olfaction, we organized the crowd-sourced DREAM Olfaction Prediction Challenge. Working from a large olfactory psychophysical dataset, teams developed machine learning algorithms to predict sensory attributes of molecules based on their chemoinformatic features. The resulting models predicted odor intensity and pleasantness with high accuracy, and also successfully predicted eight semantic descriptors (\"garlic\", \"fish\", \"sweet\", \"fruit\", \"burnt\", \"spices\", \"flower\", \"sour\"). Regularized linear models performed nearly as well as random-forest-based approaches, with a predictive accuracy that closely approaches a key theoretical limit. The models presented here make it possible to predict the perceptual qualities of virtually any molecule with an impressive degree of accuracy to reverse-engineer the smell of a molecule.\n\nOne Sentence SummaryResults of a crowdsourcing competition show that it is possible to accurately predict and reverse-engineer the smell of a molecule.

neuroscience↗