• Jenifer Velez

• Olga Lopez-Acevedo

• Cleidy Mirela Osorio-Mogollón, University of Sao Paulo, Brazil
• Leticia Cioca, University of Sao Paulo, Brazil
• Diego Leonardo, University of Sao Paulo, Brazil
• Clarice Izumi, University of Sao Paulo, Brazil
• Richard Garratt, University of Sao Paulo, Brazil
• Munira Baqui, University of Sao Paulo, Brazil

Trypanosoma brucei, the causative agent of Sleeping Sickness, is a neglected tropical disease endemic in sub-Saharan Africa. The parasite has a complex structural element, the Flagellum Attachment Zone (FAZ), which is crucial for linking the single flagellum to the cell body. Within the FAZ, high-molecular-mass proteins are essential for understanding the maintenance of cellular morphology, cytokinesis, and survival. Among these, FAZ10 is a prominent protein we have identified and characterized. FAZ10 assumes a critical function in cytokinesis, furrow positioning, and the overall organization of the FAZ. However, we have limited knowledge about the molecular structure of this intriguing protein. Our objective is to conduct a structural analysis of the central region of FAZ10, which is characterized by low intrinsic disorder and the potential presence of coiled-coil motifs. To generate FAZ10 3D models, we used AlphaFold2. Additionally, we utilized LOGICOIL, MARCOIL2, and LIGPLOT to characterize coiled-coil motifs, IUPRED2A to predict disordered regions, and PyMOL for visualizing and rendering these models. For experimental analyzes, we expressed the central region of FAZ10 through a heterologous system, and purified it using the NiNTA system, Size Exclusion Chromatography (SEC), and SEC-MALS. The structural characterization was made by circular dichroism and Cryo-EM. The central region comprises two confirmed globular domains interconnected by a coiled-coil motif, potentially facilitating protein-protein interactions. Furthermore, we postulate that FAZ10 has the capability to form a protein dimer stabilized by this motif. Expression of the central region yields a protein of ~67 kDa, while the globular domain results in a protein of ~11 kDa. Circular dichroism showed that there is a concordance between the in silico and experimental analyses of the recombinant proteins. In addition, the central region of FAZ10 is a dimer protein, which would lead us to think that the entire protein has a high potential to be dimeric. The structure is shown as a fiber approximately 40 nm in length. Subsequently, we aim to shed light on its role and interactions with other FAZ region proteins, so our forthcoming experiments will involve pulldown assays to capture proteins present in the FAZ of T. brucei in conjunction with recombinant protein expressed. These will be followed by mass spectrometry analysis. Ultimately, this will enhance our understanding of the biology of this parasite—a significant public health concern in Africa. Such insights will be crucial for identifying potential new therapeutic targets.

• Vedikaa Dhiman, Department of Bioscience and Bioengineering, Indian Institute of Technology Jodhpur, Jodhpur-342030, Rajasthan, India, India
• Soham Biswas, Department of Biotechnology and Bioinformatics, University of Hyderabad, Hyderabad, Telangana, India, India
• Rajveer Singh Shekhawat, Department of Bioscience and Bioengineering, Indian Institute of Technology Jodhpur, Jodhpur-342030, Rajasthan, India, India
• Ayan Sadhukhan, Department of Bioscience and Bioengineering, Indian Institute of Technology Jodhpur, Jodhpur- 342030, Rajasthan, India, India
• Pankaj Yadav, Department of Bioscience and Bioengineering, Indian Institute of Technology Jodhpur, Jodhpur- 342030, Rajasthan, India, India

Rice (Oryza sativa) is the major cereal crop consumed by more than half of the global population. It is a staple crop affected by biotic stress, leading to pathogenic diseases. Despite this, it possesses resistance (R) genes and pattern recognition receptor (PRR) proteins against bacterial and fungal pathogens, Xanthomonas oryzae pv. oryzae (Xoo) and Magnaporthe oryzae (M. oryzae), respectively. However, none of the rice paralogs of disease-resistance proteins has been characterized till now. In the current study, gene network analysis revealed five novel disease-resistance proteins against bacterial leaf blight (BB) and rice blast (RB) diseases. In silico modeling, refinement, and model quality assessment were performed to predict the best structures of these five proteins and submitted to ModelArchive for future use. An in silico annotation indicated that the five proteins functioned in signal transduction pathways as kinases, phospholipases, transcription factors, and DNA-modifying enzymes. The proteins were localized in the nucleus and plasma membrane. Phylogenetic analysis showed the evolutionary relation of the five proteins with disease-resistance proteins (XA21, Oryza sativa Trithorax1 (OsTrx1), Phospholipase D (PLD), and HxKxxxxDx6GSxN-motif-containing proteins). This indicates similar disease-resistant properties between five unknown proteins and their evolutionary-related proteins. Furthermore, gene expression profiling of these proteins using public microarray data showed their differential expression under Xoo and M. oryzae infection. This study provides insight into developing disease-resistant rice varieties by predicting novel candidate resistance proteins, which will assist rice breeders in improving crop yield to address future food security through molecular breeding and biotechnology.

• Gabriel Garcia, Departamento de Química Biológica, FCEN, Universidad de Buenos Aires, Argentina. IQUIBICEN-CONICET, Argentina
• German Jurado, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina., Argentina
• Miranda Palumbo, Instituto de Cálculo, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina., Argentina
• Florencia Castello, Instituto de Cálculo, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina., Argentina
• Federico Serral, Departamento de Química Biológica, FCEN, Universidad de Buenos Aires, Argentina. IQUIBICEN-CONICET, Argentina
• Ezequiel Sosa, Departamento de Sistema, Universidad Tecnologica Nacional, Buenos Aires, Argentina., Argentina
• Rafael Terra, Laboratório Nacional de Computação Científica (pt-br) National Laboratory for Scientific Computing, Petrópolis, Brazil., Brazil
• Dario Fernandez Do Porto, Instituto de Cálculo, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina., Argentina
• Adrian Turjanski, Departamento de Química Biológica, FCEN, Universidad de Buenos Aires, Argentina. IQUIBICEN-CONICET, Argentina

Diseases caused by infectious agents have been a major cause of death worldwide from ancient times to the 21st century. Currently, a multidisciplinary approach combined with advances in bioinformatics tools and sequencing techniques has led to significant progress in the development of antimicrobials. However, despite this, many resistant bacteria still result in prolonged and ineffective treatments.
In this context, we present Target Pathogen 2.0, a bioinformatics tool designed to automate the drug discovery process for potential treatments based on a bacterial genome. From this genome, all information related to sequences, structural, and functional annotations is generated and stored in a PostgreSQL database using AlphaFold and the PDB databases. Subsequently, druggable sites and the subcellular localization of proteins are identified using P2rank and PSORTb, respectively. All these platform functionalities are supported using Django, providing an online service that allows users to interact with all the information.
Additionally, we are also incorporating an original module capable of predicting, through a machine learning model, the drugs capable of interacting with the pathogen's proteins, offering researchers links to available compounds for in vitro research. Moreover, for those looking to enrich the analysis, Target Pathogen 2.0 allows users to import their own data from tools or experiments not included in the main pipeline, providing a personalized experience for the requirements of each research group.
Unlike its predecessor[1], this new version incorporates Parsl as its backbone. Parsl is a library that allows for the parallelization of Python code, and this robust manager guarantees the project's scalability and facilitates the incorporation of new modules capable of capturing the functionalities of emerging tools in the future. Target Pathogen 2.0 aims to offer a user-friendly and easily accessible environment that can be used by researchers without a background in informatics.

Sosa, E. J., Burguener, G., Lanzarotti, E., Defelipe, L., Radusky, L., Pardo, A. M., … Fernández Do Porto, D. (2018). Target-Pathogen: a structural bioinformatic approach to prioritize drug targets in pathogens. Nucleic Acids Research, 46(D1), D413–D418. https://doi.org/10.1093/nar/gkx1015

• Kush Narang, University of California, Davis, United States
• Abhigyan Nath, Pt. J.N.M Medical College, India
• William Hemstrom, Purdue University, United States
• Simon K.S. Chu, University of California, Davis, United States

The stability and function of proteins are dictated by their three-dimensional structures. It is well-known that environmental factors, such as solvent salinity, interfere with the protein folding process. However, the exact influence of salinity on protein folding is poorly understood.

Here, we present HaloClass, a machine learning algorithm that accurately predicts whether a protein will function in a high salinity environment. HaloClass establishes a new state-of-the-art, significantly outperforming all existing published literature.

HaloClass was trained by adding a classification step to the end of an ESM-2 language model. To develop HaloClass, we began by evaluating several protein language models including ProtT5 and ESM-2. We ultimately used the ESM-2 checkpoint with 650 million parameters for having the optimal balance of speed and performance. After testing multiple classification steps to attach to the end, we determined that an XGBoost tree, implemented in Python, was the ideal approach.

Our model significantly outperforms existing work on all the existing benchmark datasets in the field while having a training dataset that has less overlap with testing sets than past works. Because existing datasets in this field are small (16 and 278 proteins respectively), we created a new dataset, multiple orders of magnitude larger and more diverse than past datasets. Using this dataset containing more than 20,000 unique protein sequences from dozens of organisms, we demonstrate that HaloClass retains greater than 97 percent accuracy, compared to 78.5 percent for the existing state-of-the-art. HaloClass further shows a AUC-ROC value of 0.9941, nearly perfect, indicating that almost all salt-tolerant proteins were classified as more salt-tolerant than any non-salt-tolerant counterparts.

To gain further insight into the model’s understanding, we created figures visualizing known experimental structures from homologous proteins. Despite RMSDs between pairs of structures being as low as 0.7 angstroms, our model performs nearly perfectly, making just one mistake on the existing benchmark dataset; we analyze and explain the cause of this error in the work.

Finally, we conducted a guided, artificial mutation study where we intentionally altered surface and interior residues to demonstrate the protein folding characteristics (hydrophobic core and hydrophilic surface) that the model is able to understand.

In conclusion, our highly performant model establishes a new benchmark in the field that will progress the design and innovation of salt-tolerant enzymes. Simultaneously, the model will also help microbiologists seeking to understand the characteristics of salt-tolerant proteins.

• Luisa Mejia-Sequera, Biology and Breeding Research Program, Colombian Oil Palm Research Center, Cenipalma., Colombia
• Juan Malagón, Biology and Breeding Research Program, Colombian Oil Palm Research Center, Cenipalma., Colombia
• David Octavio Botero Rozo, Grupo de Biología y Mejoramiento, Centro de Investigaciones en Palma de Aceite-Cenipalma, Colombia., Colombia
• Hernán Mauricio Romero, Grupo de Biología y Mejoramiento, Centro de Investigaciones en Palma de Aceite-Cenipalma, Colombia., Colombia

Oil palm (Elaeis guineensis Jacq.) is an economically important crop due to the production of oil derived from its fruit. One of the most limiting factors in the oil palm crop is bud rot (BR), a disease caused by the hemibiotrophic oomycete Phytophthora palmivora, which has been affecting the palm crop in Colombia since 1964. Since 2008, the Oil Palm Research Center (Cenipalma) has led research efforts to elucidate the different mechanisms involved in the infection to generate new and efficient control strategies. Genome-Scale Metabolic Models (GSMMs) are a systems biology approach to reconstruct the metabolism of an organism and perform a mathematical analysis to calculate the mass fluxes, characterize the system and predict specific phenotypes ​(1)​. This approach uses the genome to annotate the enzymes codified for an organism, each of which catalyzes a reaction. Those reactions have stoichiometric relations. All this information can be organized on a matrix where the rows are metabolites and columns are the reactions. Using that information, we can create a metabolic network and mathematically calculate the fluxes through each of these reactions in a method called Flux Balance Analysis (FBA). Models of other species such as P. infestans have shown that the pathogen significantly changes its metabolism during the different phases of its life cycle and have revealed the importance of pathways like amino acid degradation and fructose usage ​(2,3)​.



In our study, a reconstruction of the metabolism of P. palmivora was obtained and, together with data on growth and tissue composition, an FBA was performed to elucidate the metabolic pathways with the highest flux and those important for the growth of the pathogen. Metabolic reconstruction indicates that this organism has 149 metabolic pathways associated with 5869 enzymes, and 2739 reactions, involving 2440 metabolites. Manual curation reveals that P. palmivora has a broad molecular arsenal involved in the degradation of cell wall components and many other enzymes important to the tissue colonization process. FBA shows that glycolysis and amino acid degradation are important pathways in pathogen growth. This methodology, together with the integration of omics data, seeks to increase the knowledge of pathogen-palm relationships and thus contribute to the prognosis of the behavior of bud rot in oil palm.

• Daniela Sánchez-Aristizábal, University of Antioquia, Colombia
• Juan Camilo Arboleda , Rivera, Colombia

The efficiency in the codification of DNA sequences is determining for the optimal performance of genomic analysis models. Nonetheless, conventional codification methods more often than not dismiss valuable information about the biological properties of the DNA molecules and contain only nucleotide sequence information. This deficiency is relevant because, since these inherent properties are not taken upon consideration, the results might lead to an inaccurate biological interpretation. Therefore, integrating biological properties into codification could enhance representation and analysis of said sequences, improving performance and applicability of computer models into genomics. In this research, we have evaluated a new DNA sequence codification scheme based on physico-chemical properties of the nucleotides which can potentially increase performance of Deep Learning models. In order to develop the codification, we incorporated several physico-chemical properties such as: molecular weight, polar superficial area and capability to form hydrogen bonds, among others. The objective of integrating these characteristics is to obtain more accurate and relevant properties of the DNA sequence. Subsequently, we used convolutional neural network models to assess the efficiency of this codification in two specific tasks: Splicing sites prediction and taxonomic classification. First trials have shown that our codification scheme surpasses traditional methods, such as one-hot encoding and representation of K-mer vectors, performance in both Deep Learning models. Our findings suggest that the incorporation of physico-chemical properties into the DNA sequence codification can significantly increase performance of Deep Learning models. This methodology has the potential to revolutionize the focus in bioinformatics and genomic problematics, enhancing the accuracy of the models and offering new perspectives for the integration of physico-chemical properties into biological sequences analysis. This strategy can boost performance in key areas like medical diagnosis and treatment, allowing a more precise identification of biomarkers and the creation of personalized therapy. Additionally, this methodology could impact other research fields, like functional genomics and genetic engineering, since it offers and more detailed representation of the DNA sequence and its interactions.

• Simon Villegas, Universidad Nacional de Colombia, Colombia
• Juan Pablo Hernández Ortiz, Universidad Nacional de Colombia, Colombia

The Aburrá Valley, located in Antioquia, Colombia, is an area frequently affected by episodes of air pollution that pose potential health risks to its inhabitants. This study aims to characterize the particulate matter (PM) present in the air to generate relevant information for public health decision-making. While previous research has primarily focused on the chemical components of PM, the contributions of the microbiological load within PM to health remain largely unexplored.
In this research, we investigated the composition of the air microbiome, or aerobiome, from PM2.5 samples collected in the city of Medellín during the 2022-2023 period. We analyzed microbial communities through sequencing of the complete 16S rRNA gene using Oxford Nanopore technology, coupled with taxonomic classification using Kraken2 and Bracken, as well as the Silva database (version 138). Additionally, Faprotax was employed to predict the functional profile of the aerobiome. Notably, the presence of extremophilic microorganisms was observed, with the phyla Proteobacteria, Actinobacteria, Firmicutes, and Cyanobacteria predominating. The core aerobiome comprised 49 distinct genera, which largely coincided with those previously reported as ubiquitous in the air. Both alpha and beta diversity metrics were influenced by climatic seasons.
We found that the genera Leuconostoc, Weissella, and Thermomonas were associated with moderate air quality, while the uncultivated genus 1174-901-12 from the Acetobactereaceae family and Roseomonas were differentially abundant in conditions of good air quality. Moderate air quality was associated with less robust co-occurrence networks, suggesting a disruption of the aerobiome. The functional characterization of the aerobiome revealed the presence of potentially pathogenic microorganisms and hydrocarbon degraders. Correlations were identified between aerobic nitrite oxidation, anammox, cellulolysis, lignolysis, ureolysis, sulfate and manganese oxidation, invertebrate parasites, human pathogens, and fumarate respiration with environmental variables.
This study represents the first comprehensive characterization of the aerobiome in the Aburrá Valley using a culture-independent approach, highlighting the presence of microorganisms relevant to human and environmental health, as well as potential applications in bioremediation. The findings underscore the importance of incorporating aerobiome characterization into air quality monitoring systems.

• Natalia Coutouné, Universidade Estadual de Campinas, Unicamp, Brazil

Historically, women have not been appropriately recognized in science, which is no differentin computing. Although in the beginning, the tasks related to programming and computing were carried out by women, today, we hardly reach more than 25% in organizations. Even though there is a large percentage of female students in the biological sciences, the Bioinformatics area is still predominantly male, which is why there is a need to understand and contribute to changing this reality. Thus, in 2020 PyLadies Bioinfo was born. Since then, it has brought together more than 100 self-declared women to learn, teach, and encourage through collaborative learning. In this presentation, we would like to present our initiative to promote the inclusion of women in one of the most active fields of biology today: Bioinformatics.

• Alberto Delgado

• Yanara Bernal, Instituto de Ciencias e Innovación en Medicina, Universidad del Desarrollo, Chile
• Alejandro Blanco, Instituto de Ciencias e Innovación en Medicina, Universidad del Desarrollo, Chile
• Karen Oróstica, Instituto de Investigación Interdisciplinaria, Universidad de Talca, Chile
• Ricardo Armisén, Instituto de Ciencias e Innovación en Medicina, Universidad del Desarrollo, Chile

Integrating artificial intelligence into clinical decision-making can improve cancer outcomes. RNA editing has become increasingly relevant in studying phenotypes and outcomes in cancer patients. Recently, survival prediction models have been developed based on RNA editing in gastric cancer, acute myeloid leukemia, and breast cancer (BC), as well as models predicting immunotherapy response in lung and chemotherapy in gastric cancer. We integrated classic omics, including DNA mutations and RNA gene expression, added RNA editing, and developed a drug response prediction model. We analyzed 107 patients from the Breast Cancer Genome-Guided Therapy Study (ClinicalTrials.gov: NCT02022202). This study was used to train (70%) with 10-fold cross-validation and test (30%) the drug response classification models. We used Random Forest (RF), generalized linear model (GLM), and support vector machines (SVM) with the Caret package to classify therapy response using various combinations of clinical data, mutations, gene expression, and RNA editing data. For select the final model, we compared the F1-score between models with different data combinations. First, we characterized the cohort based on clinical data, mutation landscapes, differential gene expression, and RNA-edited sites in 72 non-responders and 35 responders to therapy. Second, about the prediction models, RNA editing improved RF performance for predicting drug response across all combinations, achieving an F1 score of 0.8 (95% CI: 0.68-0.9 from bootstrapping). Finally, we identified and characterized the top ten variables of the best model, focusing on RNA-edited sites. For the selected model, we will validate our findings in two independent cohorts and calculate the actual performance metrics. RNA-edited sites could be strong predictors of drug response in BC. The improvement of the models from classical omics could be explained by the rich source of RNA for prediction and the potential involvement of RNA-edited sites in drug response based on data from cancer cell lines and patients. Our findings could be useful for the development of tests based on RNA-edited sites for predicting drug response prior to treatment, potentially improving precision medicine, survival, and quality of life for cancer patients.

• Raissa Medina Santos, Conservatoire National des Arts et Métiers, France
• Sigrid Le Clerc, Conservatoire National des Arts et Métiers, France
• Léa Bruneau, Centre Hospitalier Universitaire de La Réunion, Reunion
• Adrien Maillot, Centre Hospitalier Universitaire de La Réunion, Reunion
• Taoufik Labib, Conservatoire National des Arts et Métiers, France
• Myriam Rahmouni, Conservatoire National des Arts et Métiers, France
• Cécile Lefebvre, Vaccine Research Institute, France
• Nora El Jahrani, Vaccine Research Institute, France
• Christine Fontaine, Centre Hospitalier Universitaire de La Réunion, Reunion
• Christine Payet, Centre Hospitalier Universitaire de La Réunion, Reunion
• Nathalie Ah-You, Centre Hospitalier Universitaire de La Réunion, Reunion
• Cécile Chabert, Centre Hospitalier Universitaire de La Réunion, Reunion
• Corinne Mussard, Centre Hospitalier Universitaire de La Réunion, Reunion
• Sylvaine Porcherat, Centre Hospitalier Universitaire de La Réunion, Reunion
• Samir Medjane, Centre Hospitalier Universitaire de La Réunion, Reunion
• Josselin Noirel, Conservatoire National des Arts et Métiers, France
• Catherine Marimoutou, Centre Hospitalier Universitaire de La Réunion, Reunion
• Hakim Hocini, Vaccine Research Institute, France
• Patrick Gerardin, Centre Hospitalier Universitaire de La Réunion, Reunion
• Jean-François Zagury, Conservatoire National des Arts et Metiers, France

Background: In 2005-2006, a chikungunya epidemic of unprecedented magnitude struck La Réunion, a southwestern Indian Ocean Island, causing 300,000 infections. Over time, significant public health concerns emerged due to long-lasting manifestations, particularly chronic rheumatic and chronic fatigue-related syndromes.

Methods: To investigate the pathophysiology underlying chronic chikungunya (CC), particularly the chronic fatigue syndrome (CFS), the CHIKGene study was initiated. Blood samples were collected from 133 individuals who experienced chronic symptoms, with 58 of them presenting persistent chronic fatigue-like symptoms, 11 to 14 years after exposure. RNA-Seq was performed on purified PBMCs followed by the comparison of the mRNA gene expression profiles between CFS and CC.

Results: Gene expression analysis between CFS and CC individuals identified only three differentially expressed genes (DEGs): EGR1, EGR2, and FOSB, all down-regulated in CFS compared to CC. These genes are linked to mood disorders, schizophrenia and stress-related conditions, indicating a possible connection between psychiatric disorders and chronic chikungunya. Metascape pathway analysis revealed significant immune response-related terms predominantly overexpressed in CC. DisGeNET analysis highlighted mood disorder-related pathways, also involving ACE and PTGS2 genes, both associated with CFS.

Conclusions: This study identified key DEGs and pathways in individuals with CFS linked to CC. Down-regulation of genes associated with mood disorders and immune response alterations in CFS suggests a novel genetic correlation with the chronicity of chikungunya. These findings emphasize the importance of these genetic markers in understanding, diagnosing and potentially developing therapies for long-term chikungunya effects, particularly its impact on mental health and immune function.

• David Requena, New York University, United States
• Jack Medico, Rockefeller University, United States
• Luis Soto-Ugaldi, Rockefeller University, United States
• Mahsa Shirani, Rockefeller University, United States
• James Saltsman, Rockefeller University, United States
• Michael Torberson, Mayo Clinic, United States
• Philip Coffino, Rockefeller University, United States
• Sanford Simon, Rockefeller University, United States

Fibrolamellar Hepatocellular Carcinoma (FLC) is a rare liver cancer affecting adolescents and young adults without history of underlying viral hepatitis, cirrhosis, or other known risk factors. FLC is misclassified as a subtype of HCC, causing patients to receive HCC drug therapy, ineffective against FLC. Therefore, defining the molecular identity of FLC is critical for developing and administering proper drug treatment.

Most FLC patients present somatic heterozygous deletions in chromosome 19p13.12, causing a fusion transcript connecting exon 1 of DNAJB1 and exons 2-10 of PRKACA (the catalytic subunit of PKA). However, patients with FLC-like histopathology but different genetic alterations involving PKA have been reported.

Current FLC RNA-seq studies have limited agreement in their differentially expressed genes, attributable to the use of small datasets, lack of paired normal samples and inadequate bioinformatics methods. To better understand FLC and its relation with FLC-like and other liver tumors, we collected over 1500 samples in the largest study of liver cancer to date. These were processed using state-of-the-art bioinformatics methods, including filters of detectability, consistency and validation steps that we devised to find transcriptomic signatures unrestricted to experimental processing.

We sequenced the whole transcriptome of 127 FLC and 2 FLC-like frozen tissue samples and reprocessed 73 FLC and 18 FLC-like samples from external datasets. A transcriptomic FLC signature of 287 up- and 406 down-regulated genes was identified, which demonstrated that FLC and FLC-like tumors with diverse dysregulations of PKA are a single disease with a common transcriptome, rather than a collection of diverse diseases with similar pathologic features. We studied this signature at different omic levels. At genome, exome and methylome levels, no recurrent alterations associated with this signature were found. Interestingly, HDAC1 (which interacts with PRKACA) was found to be associated with transcription factors targeting genes of the FLC signature. Moreover, we performed spatial transcriptomics and distinguished which of the different cell types in the tumor are giving origin to the transcriptional dysregulations in the FLC signature. Primary tumors and metastases showed high similarity, with only differential expression in 0.6% of their genes. They are associated with tumor proliferation, maintenance, and immune evasion, and may help elucidate the metastatic process. We further used the FLC signature to validate organoid and PDX models. Our analysis was expanded to other liver cancers, analyzing 1192 tumor and normal samples of HCC, hepatoblastoma, and intrahepatic cholangiocarcinoma, identifying their transcriptomic signatures and comparing them with FLC.

• Gaurik Mukherjee, Chemical Engineering and Process Division, CSIR - National Chemical Laboratory (NCL), India
• Leelavati Narlikar, Department of Data Science and Department of Biology, Indian Institute of Science Education and Research Pune, India

The transcriptional state of a cell is driven in large part by the binding of specific combinations of transcription factors (TFs) to their cognate sites. These combinations of binding sites typically lie within accessible chromatin, and hence, assays that profile accessible chromatin are routinely used to understand the transcriptional regulatory state of cells. Recent technological advances in profiling chromatin accessibility at a single-cell resolution (e.g. single-cell ATAC-seq) have now enabled discovery of regulatory heterogeneity across developmental stages and tissues. Current approaches characterise this heterogeneity based on differentially accessible regions reported by the high-throughput assay. We show that regulatory diversity can be better captured by combining information from single-cell ATAC-seq experiments with transcription factor binding motifs (TFBMs). Our results demonstrate that using gain or loss of chromatin accessibility within sets of combinations of TFBMs can better explain groups of similarly behaving cells within the profiled population. Using a compendium of single-cell ATAC-seq datasets from differentiating hematopoietic cells in humans to different landmark stages in Drosophila melanogaster embryogenesis, we show that our approach can recover the different cell states and the associated TFBMs. Some of our findings validated from literature include the pivotal role of STAT-Zelda combination in Drosophila during early developmental stages and SPi1-EBF1 combination in lineage commitment of B-lymphocytes in humans. This suggests that our method has the potential to discover novel combinations of TFs and their corresponding target cell states from single cell ATAC-seq data across species.

• Sebastián Urquiza-Zurich, Advanced Center for Chronic Diseases - ACCDiS, Facultad de Ciencias Químicas y Farmacéuticas, Universidad de Chile, Chile
• Emiliano Vicencio, Advanced Center for Chronic Diseases - ACCDiS, Facultad de Ciencias Químicas y Farmacéuticas, Universidad de Chile, Chile
• Francisco Sigcho-Garrido, Advanced Center for Chronic Diseases - ACCDiS, Facultad de Ciencias Químicas y Farmacéuticas, Universidad de Chile, Chile
• Francisco Pino-de la Fuente, Advanced Center for Chronic Diseases - ACCDiS, Facultad de Ciencias Químicas y Farmacéuticas, Universidad de Chile, Chile
• Danica Jiménez-Gallegos, Advanced Center for Chronic Diseases - ACCDiS, Facultad de Ciencias Químicas y Farmacéuticas, Universidad de Chile, Chile
• Paulo Amaral, Insper Instituto de Ensino e Pesquisa, Chile
• Sergio Lavandero, Advanced Center for Chronic Diseases - ACCDiS, Facultad de Ciencias Químicas y Farmacéuticas, Universidad de Chile, Chile
• Vinicius Maracaja-Coutinho, Advanced Center for Chronic Diseases - ACCDiS, Facultad de Ciencias Químicas y Farmacéuticas, Universidad de Chile, Chile

Cardiovascular diseases (CVDs) are one of the leading causes of death worldwide and their prevalence continues to increase. They can be induced by several factors, such as neurohumoral activation (catecholamine dysregulation), and pressure overload (as in hypertension), among others. Angiotensin II (AngII), is a vasoactive peptide with a crucial role in hypertension, cardiac remodeling, and activation of different kinase pathways such as JNK and ERK. The latter is important since these pathways converge on master regulators (MRs) that control gene expression of multiple genes or determine cell fate in different contexts, especially in CVDs. It remains to be understood whether there are other mechanisms associated with neurohumoral changes mediated by other MRs using an integrative and systems biology approach in CVDs. This study aims to identify new MRs that orchestrate responses at the transcriptional level in an Ang II context, using gene co-expression modules and the generation of biological networks. We used a bulk RNA-seq dataset from mouse hearts stimulated with AngII, to prepare the data and determine gene expression, FastQC, Hisat2, featureCounts, and DESeq2 were used sequentially. The differentially expressed genes regulated by Ang II were then used to identify modular transcriptional behavior and generate specific gene regulatory networks (GRNs). For this end, we used CEMiTool and Cytoscape, respectively. This allowed us to identify TFs and MRs that could control the transcriptional changes associated with the Ang II phenotype. A total of 16.818 protein-coding genes were found, of which only 15 were up-regulated and 675 were down-regulated. In addition, four coexpression modules were obtained, in which, by an overrepresentation analysis (ORA), module 1 (M1) with 2.097 genes was found to contain Gene Ontology (GO) processes such as “Collagen containing extracellular matrix”, “Extracellular structure organization” and “Extracellular matrix structural constituent”, that are specific to the model. Through the integrated approach, 10 MRs which could be fundamental in cardiac remodeling, and 19 TFs were identified only in Ang II such as KLF8 and MYBL2. These results not only present a new layer of regulation and understanding of the molecular mechanisms underlying cardiac remodeling, but also offer potential new therapeutic targets to study CVDs, that are associated with hypertension. Data integration with a systems biology approach, allows us to unravel complex GRN and provide a deeper understanding of biological processes at the cardiac level.