Arina Surko

• Surko, Arina & Ogul, Hasan (2025). Retrieving Relevant Single-Cell RNA Sequencing Profiles with Gene Expression Clustering. In Bourbakis, Nikolaos & Filipovic, Nenad (Ed.), 2025 IEEE 25th International Conference on Bioinformatics and Bioengineering (BIBE) Proceedings. IEEE (Institute of Electrical and Electronics Engineers). ISSN 9798331558994. p. 381–386. doi: 10.1109/BIBE66822.2025.00070. Full text in Research Archive Show summary

  Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful technology for analyzing cellular heterogeneity and gene expression at high resolution. With the rapid growth of publicly available scRNA-seq databases, there is a growing demand for methods that enable systematic comparison and retrieval of profiles based on gene expression patterns rather than metadata. However, no standardized workflow currently exists for retrieving relevant scRNA-seq profiles using expressionbased similarity. In this paper, we present a novel framework for comparing and retrieving scRNA-seq profiles based on gene expression. Each profile is represented as a dendrogram derived from hierarchical clustering of genes based on their expression patterns across cells. To compare profiles, we introduce a new similarity metric, Multi-Resolution Adjusted Rand Index (MRARI), which extends the traditional Adjusted Rand Index (ARI) by incorporating hierarchical information. Retrieval is performed by computing pairwise MR-ARI scores between profiles. We evaluate the method's performance using a bootstrapped distribution of the Area Under the Receiver Operating Characteristic Curve (AUROC).

View all works in NVA

• Surko, Arina & Ogul, Hasan (2025). Retrieving Relevant Single-Cell RNA Sequencing Profiles with Gene Expression Clustering. doi: https:/easyconferences.eu/bibe2025/detailed-program/. Full text in Research Archive Show summary

  Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful technology for analyzing cellular heterogeneity and gene expression at high resolution. With the rapid growth of publicly available scRNA-seq databases, there is a growing demand for methods that enable systematic comparison and retrieval of profiles based on gene expression patterns rather than metadata. However, no standardized workflow currently exists for retrieving relevant scRNA-seq profiles using expressionbased similarity. In this paper, we present a novel framework for comparing and retrieving scRNA-seq profiles based on gene expression. Each profile is represented as a dendrogram derived from hierarchical clustering of genes based on their expression patterns across cells. To compare profiles, we introduce a new similarity metric, Multi-Resolution Adjusted Rand Index (MRARI), which extends the traditional Adjusted Rand Index (ARI) by incorporating hierarchical information. Retrieval is performed by computing pairwise MR-ARI scores between profiles. We evaluate the method's performance using a bootstrapped distribution of the Area Under the Receiver Operating Characteristic Curve (AUROC).

• Surko, Arina & Ogul, Hasan (2025). Transfer retrieval: A new paradigm for person-alised medicine through AI-driven genomic database search. doi: 10.5617/nmi.12551. Full text in Research Archive Show summary

  Genetic diseases can differ significantly among patientsdue to the variations in the human genome. Personalisedmedicine aims to facilitate the treatment’s efficacyby tailoring it to each patient based on their uniquegenetic profile. Transfer retrieval is a novel concept forpersonalised medicine, based on assessing the similaritybetween patients’ profiles to enable the content-basedsearch of relevant genomic profiles in large, clinicallyannotated databases. A challenge in developing transferretrieval is the presence of a variety of technologiesand data modalities of gene expression profiles. Thisdiversity complicates the similarity assessment betweenthe profiles. To address this issue, transfer retrieval willenable the similarity assessment by transforming eachraw measurement data into a common and biologicallyinterpretable concept. This approach will increasethe possibilities of transfer retrieval and, also, keep itagile for future discoveries. More advanced sequencingtechnologies are emerging, such as single-cell RNAsequencing (sc-RNAseq), which allows analysing each cellfrom the sample individually. Currently, we are buildingand testing the transfer retrieval models for the sc-RNAseq data. As for today, two transfer retrieval modelsare tested. The first model was built as a baseline, and it isbased on the average expression per gene across all cells ofthe sample, imitating bulk RNAseq data. The second oneis based on the differential expression data for a cancerdataset. The tested distance metrics include Euclideanand MSE distances, Cosine and Jaccard similarities. Theretrieval results are assessed via the Area Under Curve(AUC) metric.

• Surko, Arina & Ogul, Hasan (2025). Transfer Retrieval: A new paradigm for personalised medicine through AI-driven genomic database search. doi: https:/mmiv.no/wp-content/uploads/2025/11/PRESIMAL_Abstract_Book_2025.pdf. Full text in Research Archive
• Surko, Arina & Ogul, Hasan (2025). Retrieving Relevant Single-Cell RNA Sequencing Profiles with Gene Expression Clustering. doi: https:/www.nora.ai/resources/pdf-files/naim_2025_poster_sessions_abstracts.pdf. Full text in Research Archive Show summary

  Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful technology for analyzing cellular heterogeneity and gene expression at high resolution. With the rapid growth of publicly available scRNA-seq databases, there is a growing demand for methods that enable systematic comparison and retrieval of profiles based on gene expression patterns rather than metadata. However, no standardized workflow currently exists for retrieving relevant scRNA-seq profiles using expression-based similarity. In this work, we present a novel framework for comparing and retrieving scRNA-seq profiles based on gene expression. Each profile is represented as a dendrogram derived from hierarchical clustering of genes based on their expression patterns across cells. To compare profiles, we introduce a new similarity metric, Multi-Resolution Adjusted Rand Index (MR-ARI), which extends the traditional Adjusted Rand Index (ARI) by incorporating hierarchical information. Retrieval is performed by computing pairwise MR-ARI scores between profiles. We evaluate the method's performance using a bootstrapped distribution of the Area Under the Receiver Operating Characteristic Curve (AUROC).

• Halnes, Geir; Javaid, Anam; Solvang, Michael; Nichele, Stefano; Riegler, Michael & Singstad, Bjørn-Jostein [Show all 82 contributors for this article] (2025). Proceedings of NORA’s annual conference 2025. Nordic Machine Intelligence (NMI). 5(3), p. 1–14. doi: 10.5617/nmi.12551. Full text in Research Archive

View all works in NVA