Biology subjects

Inecik, K.

Publications and source records attributed to Inecik, K..

4 recordsLinked to original sources

bioRxiv2023

flowVI: Flow Cytometry Variational Inference

Single-cell flow cytometry stands as a pivotal instrument in both biomedical research and clinical practice, not only offering invaluable insights into cellular phenotypes and functions but also significantly advancing our understanding of various patient states. However, its potential is often constrained by factors such as technical limitations, noise interference, and batch effects, which complicate comparison between flow cytometry experiments and compromise its overall impact. Recent advances in deep representation learning have demonstrated promise in overcoming similar challenges in related fields, particularly in the context of single-cell transcriptomic sequencing data analysis. Here, we propose flowVI, a multimodal deep generative model, tailored for integrative analysis of multiple massively parallel cytometry datasets from diverse sources. By effectively modeling noise variances, technical biases, and batch-specific heterogeneity using probabilistic data representation, we demonstrate that flowVI not only excels in the imputation of missing protein markers but also seamlessly integrates data from distinct cytometry panels. FlowVI thus emerges as a potent tool for constructing comprehensive flow cytometry atlases and enhancing the precision of flow cytometry data analyses. The source code for replicating these findings is hosted on GitHub, theislab/flowVI

bioinformatics↗

bioRxiv2023

scARE: Attribution Regularization for Single Cell Representation Learning

Single-cell data generation techniques have provided valuable insights into the intricate nature of cellular heterogeneity. However, effectively unraveling subtle variations within a specific gene set of interest, while mitigating the confounding presence of higher-order variability, remains challenging. To address this, we propose scARE, a novel end-to-end generative deep learning model, amplifies model sensitivity to a preselected subset of features while minimizing others. scARE incorporates an auxiliary attribution loss term during model training, which empowers researchers to manipulate the models behavior robustly and flexibly. In this study, we showcased scAREs applicability in two concrete scenarios: uncovering subclusters associated with the expression patterns of two cellular pathway genes, and its ability to optimize the model training procedure by leveraging time-points metadata, resulting in improved downstream performance.

bioinformatics↗

bioRxiv2022

High-throughput profiling of drug interactions in Gram-positive bacteria

Drug combinations present a powerful strategy to tackle antimicrobial resistance, but have not been systematically tested in many bacterial species. Here, we used an automated high-throughput setup to profile [~] 8000 combinations between 65 antibacterial drugs in three Gram-positive species: the model species, Bacillus subtilis and two prominent pathogens, Staphylococcus aureus and Streptococcus pneumoniae. Thereby, we recapitulate previously known drug interactions, but also identify ten times more interactions than previously reported in the pathogen S. aureus, including two synergies that were also effective in multi-drug resistant clinical S. aureus isolates in vitro and in vivo. Interactions were largely species-specific and mostly synergistic for drugs targeting the same cellular process, as observed also for Gram-negative species1. Yet, the dominating synergies are clearly distinct between Gram-negative and Gram-positive species, and are driven by different bottlenecks in drug uptake and vulnerabilities of their cell surface structures. To further explore interactions of commonly prescribed non-antibiotic drugs with antibiotics, we tested 2728 of such combinations in S. aureus, detecting a plethora of unexpected antagonisms that could compromise the efficacy of antimicrobial treatments in the age of polypharmacy. We uncovered even more synergies than antagonisms, some of which we could demonstrate as effective combinations in vivo against multi-drug resistant clinical isolates. Among them, we showed that the antiaggregant ticagrelor interferes with purine metabolism and changes the surface charge of S. aureus, leading to strong synergies with cationic antibiotics. Overall, this exemplifies the untapped potential of approved non-antibacterial drugs to be repurposed as antibiotic adjuvants. All data can be browsed through an interactive interface (https://apps.embl.de/combact/).

microbiology↗

bioRxiv2022

MultiCPA: Multimodal Compositional Perturbation Autoencoder

Single-cell multimodal profiling provides a high-resolution view of cellular information. Recently, multimodal profiling approaches have been coupled with CRISPR technologies to perform pooled screens of single or combinatorial perturbations. This opens the possibility of exploring the massive space of combinatorial perturbations and their regulatory effects computationally from the extrapolation of a few experimentally feasible combinations. Here, we propose MultiCPA, an end-to-end generative architecture to predict multimodal perturbation response at single cell level. Two mixing strategies to integrate multiple modalities are introduced and compared with existing methods. MultiCPA was also shown to accurately predict unseen combinatorial perturbation responses for multiple modalities. The code to reproduce the results is available on GitHub, theislab/multicpa.

systems biology↗