TY - JOUR AU - Grexa, István AU - Iván, Zsanett Zsófia AU - Migh, Ede AU - Kovács, Ferenc AU - Bolck, Hella A AU - Zheng, Xiang AU - Mund, Andreas AU - Moshkov, Nikita AU - Csapóné Miczán, Vivien AU - Koós, Krisztián AU - Horváth, Péter TI - SuperCUT, an unsupervised multimodal image registration with deep learning for biomedical microscopy JF - BRIEFINGS IN BIOINFORMATICS J2 - BRIEF BIOINFORM VL - 25 PY - 2024 IS - 2 PG - 11 SN - 1467-5463 DO - 10.1093/bib/bbae029 UR - https://m2.mtmt.hu/api/publication/34786803 ID - 34786803 AB - Numerous imaging techniques are available for observing and interrogating biological samples, and several of them can be used consecutively to enable correlative analysis of different image modalities with varying resolutions and the inclusion of structural or molecular information. Achieving accurate registration of multimodal images is essential for the correlative analysis process, but it remains a challenging computer vision task with no widely accepted solution. Moreover, supervised registration methods require annotated data produced by experts, which is limited. To address this challenge, we propose a general unsupervised pipeline for multimodal image registration using deep learning. We provide a comprehensive evaluation of the proposed pipeline versus the current state-of-the-art image registration and style transfer methods on four types of biological problems utilizing different microscopy modalities. We found that style transfer of modality domains paired with fully unsupervised training leads to comparable image registration accuracy to supervised methods and, most importantly, does not require human intervention. LA - English DB - MTMT ER - TY - JOUR AU - Moshkov, Nikita AU - Bornholdt, Michael AU - Benoit, Santiago AU - Smith, Matthew AU - Mcquin, Claire AU - Goodman, Allen AU - Senft, Rebecca A. AU - Han, Yu AU - Babadi, Mehrtash AU - Horváth, Péter AU - Cimini, Beth A. AU - Carpenter, Anne E. AU - Singh, Shantanu AU - Caicedo, Juan C. TI - Learning representations for image-based profiling of perturbations JF - NATURE COMMUNICATIONS J2 - NAT COMMUN VL - 15 PY - 2024 IS - 1 PG - 17 SN - 2041-1723 DO - 10.1038/s41467-024-45999-1 UR - https://m2.mtmt.hu/api/publication/34782532 ID - 34782532 N1 - Funding Agency and Grant Number: ECOST action [CA15124]; LENDULET-BIOMAG Grant [2018-342]; European Regional Development Funds [GINOP-2.2.1-15-2017-00072]; H2020; EU-Horizont (ERAPERMED-COMPASS, ERAPERMED-SYMMETRY, DiscovAIR, FAIR-CHARM, SWEEPICS); ELKH-Excellence grants; Cooperative Doctoral Programme (2020-2021) of the Ministry for Innovation and Technology; OTKA-SNN [P41 GM135019]; NIH [2018-192059]; Chan Zuckerberg Initiative DAF [2348683]; Silicon Valley Community Foundation; Schmidt Fellowship program of the Broad Institute; NSF DBI Award [R35 GM122547]; [TKP2021-EGA09]; [139455/ARRS]; [2020-225720] Funding text: We thank Salil Bhate for the valuable discussions and feedback provided to improve the clarity of this manuscript. NM acknowledges the short-term scientific mission grants provided by eCOST action CA15124 (NEUBIAS) in 2019 and 2020. NM and PH acknowledge support from the LENDULET-BIOMAG Grant (2018-342), from the European Regional Development Funds (GINOP-2.2.1-15-2017-00072), from the H2020 and EU-Horizont (ERAPERMED-COMPASS, ERAPERMED-SYMMETRY, DiscovAIR, FAIR-CHARM, SWEEPICS), from TKP2021-EGA09, from the ELKH-Excellence grants and from the Cooperative Doctoral Programme (2020-2021) of the Ministry for Innovation and Technology, from OTKA-SNN no.139455/ARRS. Researchers in the Carpenter-Singh lab were supported by NIH R35 GM122547 to AEC. Researchers in the Cimini lab were supported by NIH P41 GM135019. AG was supported by grant number 2018-192059 and BAC was additionally supported by grant number 2020-225720 from the Chan Zuckerberg Initiative DAF, an advised fund of the Silicon Valley Community Foundation. JCC was supported by the Schmidt Fellowship program of the Broad Institute and by the NSF DBI Award 2348683. AB - Measuring the phenotypic effect of treatments on cells through imaging assays is an efficient and powerful way of studying cell biology, and requires computational methods for transforming images into quantitative data. Here, we present an improved strategy for learning representations of treatment effects from high-throughput imaging, following a causal interpretation. We use weakly supervised learning for modeling associations between images and treatments, and show that it encodes both confounding factors and phenotypic features in the learned representation. To facilitate their separation, we constructed a large training dataset with images from five different studies to maximize experimental diversity, following insights from our causal analysis. Training a model with this dataset successfully improves downstream performance, and produces a reusable convolutional network for image-based profiling, which we call Cell Painting CNN. We evaluated our strategy on three publicly available Cell Painting datasets, and observed that the Cell Painting CNN improves performance in downstream analysis up to 30% with respect to classical features, while also being more computationally efficient. Assessing cell phenotypes in image-based assays requires solid computational methods for transforming images into quantitative data. Here, the authors present a strategy for learning representations of treatment effects from high-throughput imaging, following a causal interpretation. LA - English DB - MTMT ER - TY - JOUR AU - Moshkov, Nikita AU - Becker, Tim AU - Yang, Kevin AU - Horváth, Péter AU - Dancik, Vlado AU - Wagner, Bridget K. AU - Clemons, Paul A. AU - Singh, Shantanu AU - Carpenter, Anne E. AU - Caicedo, Juan C. TI - Predicting compound activity from phenotypic profiles and chemical structures JF - NATURE COMMUNICATIONS J2 - NAT COMMUN VL - 14 PY - 2023 IS - 1 SN - 2041-1723 DO - 10.1038/s41467-023-37570-1 UR - https://m2.mtmt.hu/api/publication/33770770 ID - 33770770 N1 - Broad Institute of MIT and Harvard, Cambridge, United States Biological Research Centre, Szeged, Hungary University of California, Berkeley, United States Cited By :1 Export Date: 19 June 2023 Correspondence Address: Caicedo, J.C.; Broad Institute of MIT and HarvardUnited States; email: jcaicedo@broad.mit.edu AB - Predicting assay results for compounds virtually using chemical structures and phenotypic profiles has the potential to reduce the time and resources of screens for drug discovery. Here, we evaluate the relative strength of three high-throughput data sources—chemical structures, imaging (Cell Painting), and gene-expression profiles (L1000)—to predict compound bioactivity using a historical collection of 16,170 compounds tested in 270 assays for a total of 585,439 readouts. All three data modalities can predict compound activity for 6–10% of assays, and in combination they predict 21% of assays with high accuracy, which is a 2 to 3 times higher success rate than using a single modality alone. In practice, the accuracy of predictors could be lower and still be useful, increasing the assays that can be predicted from 37% with chemical structures alone up to 64% when combined with phenotypic data. Our study shows that unbiased phenotypic profiling can be leveraged to enhance compound bioactivity prediction to accelerate the early stages of the drug-discovery process. LA - English DB - MTMT ER - TY - THES AU - Moshkov, Nikita TI - Application of deep learning algorithms to single-cell segmentation and phenotypic profiling PB - Szegedi Tudományegyetem PY - 2022 SP - 69 DO - 10.14232/phd.11423 UR - https://m2.mtmt.hu/api/publication/34118894 ID - 34118894 N1 - 3. Témavezető: Juan C. Caicedo, Ph.D. Broad Institut LA - English DB - MTMT ER - TY - JOUR AU - Hollandi, Réka AU - Moshkov, Nikita AU - Paavolainen, Lassi AU - Tasnádi, Ervin Áron AU - Piccinini, Filippo AU - Horváth, Péter TI - Nucleus segmentation: towards automated solutions JF - TRENDS IN CELL BIOLOGY J2 - TRENDS CELL BIOL VL - 32 PY - 2022 IS - 4 SP - 295 EP - 310 PG - 16 SN - 0962-8924 DO - 10.1016/j.tcb.2021.12.004 UR - https://m2.mtmt.hu/api/publication/32751617 ID - 32751617 N1 - Synthetic and Systems Biology Unit, Biological Research Centre (BRC), H-6726, Szeged, Hungary Doctoral School of Interdisciplinary Medicine, University of Szeged, Szeged, Hungary Laboratory on AI for Computational Biology, Faculty of Computer Science, HSE University, Moscow, Russian Federation Institute for Molecular Medicine Finland (FIMM), HiLIFE, University of Helsinki, Helsinki, FI-00014, Finland Doctoral School of Computer Science, University of Szeged, Szeged, Hungary IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST) “Dino Amadori”, I-47014 Meldola (FC), Italy Single-Cell Technologies Ltd., H-6726, Szeged, Hungary Export Date: 12 July 2022 CODEN: TCBIE Correspondence Address: Horvath, P.Synthetic and Systems Biology Unit, Biological Research Centre (BRC), H-6726, Hungary; email: horvath.peter@brc.hu LA - English DB - MTMT ER - TY - JOUR AU - Moshkov, Nikita AU - Smetanin, Aleksandr AU - Tatarinova, Tatiana V TI - Local ancestry prediction with PyLAE JF - PEERJ J2 - PEERJ VL - 9 PY - 2021 PG - 24 SN - 2167-8359 DO - 10.7717/peerj.12502 UR - https://m2.mtmt.hu/api/publication/32557475 ID - 32557475 AB - We developed PyLAE, a new tool for determining local ancestry along a genome using whole-genome sequencing data or high-density genotyping experiments. PyLAE can process an arbitrarily large number of ancestral populations (with or without an informative prior). Since PyLAE does not involve estimating many parameters, it can process thousands of genomes within a day. PyLAE can run on phased or unphased genomic data. We have shown how PyLAE can be applied to the identification of differentially enriched pathways between populations. The local ancestry approach results in higher enrichment scores compared to whole-genome approaches. We benchmarked PyLAE using the 1000 Genomes dataset, comparing the aggregated predictions with the global admixture results and the current gold standard program RFMix. Computational efficiency, minimal requirements for data pre-processing, straightforward presentation of results, and ease of installation make PyLAE a valuable tool to study admixed populations. LA - English DB - MTMT ER - TY - JOUR AU - Grexa, István AU - Diósdi, Ákos AU - Harmati, Mária AU - Kriston, András AU - Moshkov, Nikita AU - Buzás, Krisztina AU - Pietiäinen, Vilja AU - Koós, Krisztián AU - Horváth, Péter TI - SpheroidPicker for automated 3D cell culture manipulation using deep learning JF - SCIENTIFIC REPORTS J2 - SCI REP VL - 11 PY - 2021 IS - 1 PG - 11 SN - 2045-2322 DO - 10.1038/s41598-021-94217-1 UR - https://m2.mtmt.hu/api/publication/32129396 ID - 32129396 N1 - Funding Agency and Grant Number: ATTRACT project by EC [777222]; European Regional Development FundsEuropean Commission [GINOP-2.3.2-15-2016-00006, GINOP-2.3.2-15-2016-00026, GINOP-2.3.2-15-2016-00037]; Chan Zuckerberg Initiative; Cancer Foundation Finland; Academy of FinlandAcademy of FinlandEuropean Commission [337036]; LENDULET-BIOMAG Grant [2018-342]; H2020-discovAIR [874656]; Seed Networks for the HCA-DVP; NVidia GPU Grant program; [EFOP 3.6.3-VEKOP-16-2017-00009] Funding text: This project has received funding from the ATTRACT project funded by the EC under Grant Agreement 777222. We acknowledge support from the LENDULET-BIOMAG Grant (2018-342), from the European Regional Development Funds (GINOP-2.3.2-15-2016-00006, GINOP-2.3.2-15-2016-00026, GINOP-2.3.2-15-2016-00037), from H2020-discovAIR (874656), and from Chan Zuckerberg Initiative, Seed Networks for the HCA-DVP, and Cancer Foundation Finland. We acknowledge support from the EFOP 3.6.3-VEKOP-16-2017-00009 grant. The authors would like to thank Ji Ming Wang (National Cancer Institute-Frederick, Frederick, MD, USA) for kindly providing us with the 5-8F cell line. We thank the NVidia GPU Grant program for providing a Titan Xp. We also acknowledge the support (337036 grant decision) from Academy of Finland for the establishment of the Single Cell Competence Center (FIMM, HiLIFE, UH, Finland) with Biocenter Finland and FIMM High Content Imaging and Analysis unit (HiLIFE, UH and Biocenter Finland). LA - English DB - MTMT ER - TY - JOUR AU - Kornienko, I. V AU - Faleeva, T. G. AU - Schurr, T. G. AU - Aramova, O. Yu AU - Ochir-Goryaeva, M. A. AU - Batieva, E. F. AU - Vdovchenkov, E. V. AU - Moshkov, Nikita AU - Kukanova, V. V. AU - Ivanov, I. N. AU - Sidorenko, Yu S. AU - Tatarinova, T. V TI - Y-Chromosome Haplogroup Diversity in Khazar Burials from Southern Russia JF - RUSSIAN JOURNAL OF GENETICS J2 - RUSS J GENET+ VL - 57 PY - 2021 IS - 4 SP - 477 EP - 488 PG - 12 SN - 1022-7954 DO - 10.1134/S1022795421040049 UR - https://m2.mtmt.hu/api/publication/32048188 ID - 32048188 AB - Genetic studies of archaeological burials open up new possibilities for investigating the cultural-historical development of ancient populations, providing objective data that can be used to investigate the most controversial problems of archeology. In this work, we analyzed the Y-chromosomes of nine skeletons recovered from elite burial mounds attributed to the 7th-9th centuries of the Khazar Khaganate in the modern Rostov region. Genotyping of polymorphic microsatellite loci of the Y chromosome made it possible to establish that among the nine skeletons studied, three individuals had R1a Y-haplogroup, two had C2b, and one each had G2a, N1a, Q, and R1b Y-haplogroups. Such results were noteworthy for the mixture of West Eurasian and East Asian paternal lineages in these samples. The Y-chromosome data are consistent with the results of the craniological study and genome-wide analysis of the same individuals in showing mixed genetic origins for the early medieval Khazar nobility. These findings are not surprising in light of the history of the Khazar Khaganate, which arose through its separation from the Western Turkic Khaganate and establishment in the North Caucasus and East European steppes. LA - English DB - MTMT ER - TY - JOUR AU - Hollandi, Réka AU - Diósdi, Ákos AU - Hollandi, Gábor AU - Moshkov, Nikita AU - Horváth, Péter TI - AnnotatorJ: an ImageJ plugin to ease hand annotation of cellular compartments JF - MOLECULAR BIOLOGY OF THE CELL J2 - MOL BIOL CELL VL - 31 PY - 2020 IS - 20 SP - 2179 EP - 2186 PG - 8 SN - 1059-1524 DO - 10.1091/mbc.E20-02-0156 UR - https://m2.mtmt.hu/api/publication/31612924 ID - 31612924 N1 - Cited By :13 Export Date: 12 July 2022 CODEN: MBCEE Correspondence Address: Horváth, P.; Synthetic and Systems Biology Unit, Hungary; email: horvath.peter@brc.hu AB - AnnotatorJ combines single-cell identification with deep learning (DL) and manual annotation. Cellular analysis quality depends on accurate and reliable detection and segmentation of cells so that the subsequent steps of analyses, for example, expression measurements, may be carried out precisely and without bias. DL has recently become a popular way of segmenting cells, performing unimaginably better than conventional methods. However, such DL applications may be trained on a large amount of annotated data to be able to match the highest expectations. High-quality annotations are unfortunately expensive as they require field experts to create them, and often cannot be shared outside the lab due to medical regulations. We propose AnnotatorJ, an ImageJ plugin for the semiautomatic annotation of cells (or generally, objects of interest) on (not only) microscopy images in 2D that helps find the true contour of individual objects by applying U-Net-based presegmentation. The manual labor of hand annotating cells can be significantly accelerated by using our tool. Thus, it enables users to create such datasets that could potentially increase the accuracy of state-of-the-art solutions, DL or otherwise, when used as training data. LA - English DB - MTMT ER - TY - JOUR AU - Moshkov, Nikita AU - Mathe, Botond AU - Kertész-Farkas, Attila AU - Hollandi, Réka AU - Horváth, Péter TI - Test-time augmentation for deep learning-based cell segmentation on microscopy images JF - SCIENTIFIC REPORTS J2 - SCI REP VL - 10 PY - 2020 IS - 1 PG - 7 SN - 2045-2322 DO - 10.1038/s41598-020-61808-3 UR - https://m2.mtmt.hu/api/publication/31595768 ID - 31595768 AB - Recent advancements in deep learning have revolutionized the way microscopy images of cells are processed. Deep learning network architectures have a large number of parameters, thus, in order to reach high accuracy, they require a massive amount of annotated data. A common way of improving accuracy builds on the artificial increase of the training set by using different augmentation techniques. A less common way relies on test-time augmentation (TTA) which yields transformed versions of the image for prediction and the results are merged. In this paper we describe how we have incorporated the test-time argumentation prediction method into two major segmentation approaches utilized in the single-cell analysis of microscopy images. These approaches are semantic segmentation based on the U-Net, and instance segmentation based on the Mask R-CNN models. Our findings show that even if only simple test-time augmentations (such as rotation or flipping and proper merging methods) are applied, TTA can significantly improve prediction accuracy. We have utilized images of tissue and cell cultures from the Data Science Bowl (DSB) 2018 nuclei segmentation competition and other sources. Additionally, boosting the highest-scoring method of the DSB with TTA, we could further improve prediction accuracy, and our method has reached an ever-best score at the DSB. LA - English DB - MTMT ER -