MicroRNAs (miRNAs) are small non coding RNAs essentially known to repress the expression of protein coding genes, either by degrading mRNAs or by preventing their translation into proteins by binding to the mRNA 3’ UTR [1], even if some cases of gene upregulation have also been documented [2]. While many available tools allow to predict miRNA genes genomewide, the identification of their targets is still a challenging task, that relies most of the time on error prone sequence based predictions. Here we aim at making use of RNA-seq of small and long RNAs in a set of chicken samples (7 tissues, 3 developmental stages and 4 animals, GENE-SWitCH project, https://www.gene-switch.eu/) in order to identify both the set of miRNAs that are expressed in these samples, and their putative target protein coding genes. We also see this approach as a step towards a flexible and incremental framework to address genomic regulation in multi-omics datasets, that allows an intuitive representation of data that can be expanded along the project. Here we decided to use knowledge graphs where nodes are biological objects such as miRNAs, genes or regulatory genomic elements, and where edges represent different kinds of evidence of regulatory relationships between biological objects and collected either from pre-processing of sequencing or genome data. More specifically we have used neo4j and its querying tool cypher (http://neo4j.com) to store simple genomic relations between miRNAs and genes, and then retrieve a set of reliable miRNA/gene relationships from them by querying the graph. More precisely we have used as our initial set of miRNAs the 2249 mirdeep2 miRNAs identified in at least 2 of our 84 chicken samples (RNA-seq of small RNAs, [3]). As for long genes we used the 33,932 genes identified from the same 84 chicken samples using the TAGADA pipeline (RNA-seq of long RNAs, [4]). We then computed three types of relations between miRNAs and genes: 1) correlation of expression between miRNAs and genes, 2) prediction of miRNA binding site on the 3’ UTR of mRNAs using TargetScan [ref], 3) Inference score computed by the Genie3 software [ref], using miRNAs as predictors of the transcriptome expression. The resulting knowledge graph was made of 410,604 nodes and 162,056,558 edges. Using Cypher, we extracted 71,726 miRNA-mRNA pairs exhibiting a good binding site and for which Genie3 considered the miRNA as being a potential regulator of the mRNA. Surprisingly, 73.9% of those relations were actually positive correlations between miRNA and mRNA expression. Restricting our analysis to negative correlations and using orthology with the human genome, we recovered respectively 70 and 134 of the relationships present in the reference MirTarBase [7] and Diana-tarbase [8] databases. In the case of MirTarBase, it represents a slight enrichment over the results obtained starting from all possible initial miRNA-gene relationships (exact Fisher test, p-value=0.08). Our study shows it is possible to use a knowledge graph to store miRNA/gene relationships obtained from different methods and to interrogate it to obtain a meaningful set of relations, even in the case of a less studied species such as chicken. Adding other kinds of regulatory relationships and finding patterns in this graph, could help us uncover new biology related to chicken development. References 1. Shang R, Lee S, Senavirathne G, Lai EC. microRNAs in action: biogenesis, function and regulation. Nature Reviews Genetics. 2023 Dec;24(12):816-33. 2. Valinezhad Orang A, Safaralizadeh R, Kazemzadeh-Bavili M. Mechanisms of miRNA-mediated gene regulation from common downregulation to mRNA-specific upregulation. International journal of genomics. 2014 Oct;2014. 3. Friedländer MR, Mackowiak SD, Li N, Chen W, Rajewsky N. miRDeep2 accurately identifies known and hundreds of novel microRNA genes in seven animal clades. Nucleic acids research. 2012 Jan 1;40(1):37-52. 4. Kurylo C, Guyomar C, Foissac S, Djebali S. TAGADA: a scalable pipeline to improve genome annotations with RNA-seq data. NAR Genomics and Bioinformatics. 2023 Dec 1;5(4):lqad089. 5. Lewis BP, Burge CB, Bartel DP. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. cell. 2005;120(1):15‑20. 6. Huynh-Thu VA, Irrthum A, Wehenkel L, Geurts P. Inferring regulatory networks from expression data using tree-based methods. PloS one. 2010;5(9):e12776. 7. https://mirtarbase.cuhk.edu.cn/~miRTarBase/miRTarBase_2022/php/index.php 8. https://dianalab.e-ce.uth.gr/tarbasev9