Motivation Drug-target interaction (DTI) prediction is a relevant but challenging task in the drug repurposing field. In-silico approaches have drawn particular attention as they can reduce associated costs and time commitment of traditional methodologies. Yet, current state-of-the-art methods present several limitations: existing DTI prediction approaches are computationally expensive, thereby hindering the ability to use large networks and exploit available datasets and, the generalization to unseen datasets of DTI prediction methods remains unexplored, which could potentially improve the development processes of DTI inferring approaches in terms of accuracy and robustness.Results In this work, we introduce GeNNius (Graph Embedding Neural Network Interaction Uncovering System), a Graph Neural Network (GNN)-based method that outperforms state-of-the-art models in terms of both accuracy and time efficiency across a variety of datasets. We also demonstrated its prediction power to uncover new interactions by evaluating not previously known DTIs for each dataset. We further assessed the generalization capability of GeNNius by training and testing it on different datasets, showing that this framework can potentially improve the DTI prediction task by training on large datasets and testing on smaller ones. Finally, we investigated qualitatively the embeddings generated by GeNNius, revealing that the GNN encoder maintains biological information after the graph convolutions while diffusing this information through nodes, eventually distinguishing protein families in the node embedding space.

Survival rates of patients with metastatic castration-resistant prostate cancer (mCRPC) are low due to lack of response or acquired resistance to available therapies, such as abiraterone (Abi). A better understanding of the underlying molecular mechanisms is needed to identify effective targets to overcome resistance. Given the complexity of the transcriptional dynamics in cells, differential gene expression analysis of bulk transcriptomics data cannot provide sufficient detailed insights into resistance mechanisms. Incorporating network structures could overcome this limitation to provide a global and functional perspective of Abi resistance in mCRPC. Here, we developed TraRe, a computational method using sparse Bayesian models to examine phenotypically driven transcriptional mechanistic differences at three distinct levels: transcriptional networks, specific regulons, and individual transcription factors (TF). TraRe was applied to transcriptomic data from 46 patients with mCRPC with Abi-response clinical data and uncovered abrogated immune response transcriptional modules that showed strong differential regulation in Abi-responsive compared with Abi-resistant patients. These modules were replicated in an independent mCRPC study. Furthermore, key rewiring predictions and their associated TFs were experimentally validated in two prostate cancer cell lines with different Abi-resistance features. Among them, ELK3, MXD1, and MYB played a differential role in cell survival in Abi-sensitive and Abi-resistant cells. Moreover, ELK3 regulated cell migration capacity, which could have a direct impact on mCRPC. Collectively, these findings shed light on the underlying transcriptional mechanisms driving Abi response, demonstrating that TraRe is a promising tool for generating novel hypotheses based on identified transcriptional network disruptions.

Despite the abundance of multimodal data, suitable statistical models that can improve our understanding of diseases with genetic underpinnings are challenging to develop. Here, we present SparseGMM, a statistical approach for gene regulatory network discovery. SparseGMM uses latent variable modeling with sparsity constraints to learn Gaussian mixtures from multiomic data. By combining coexpression patterns with a Bayesian framework, SparseGMM quantitatively measures confidence in regulators and uncertainty in target gene assignment by computing gene entropy. We apply SparseGMM to liver cancer and normal liver tissue data and evaluate discovered gene modules in an independent single-cell RNA sequencing (scRNA-seq) dataset. SparseGMM identifies PROCR as a regulator of angiogenesis and PDCD1LG2 and HNF4A as regu-lators of immune response and blood coagulation in cancer. Furthermore, we show that more genes have significantly higher entropy in cancer compared with normal liver. Among high-entropy genes are key multifunctional components shared by critical pathways, including p53 and estrogen signaling.

Inflammation is a common feature in neurodegenerative diseases that contributes to neuronal loss. Previously, we demonstrated that the basal inflammatory tone differed between brain regions and, consequently, the reaction generated to a pro-inflammatory stimulus was different. In this study, we assessed the innate immune reaction in the midbrain and in the striatum using an experimental model of Parkinson's disease. An adeno-associated virus serotype 9 expressing the alpha-synuclein and mCherry genes or the mCherry gene was administered into the substantia nigra. Myeloid cells (CD11b(+)) and astrocytes (ACSA2(+)) were purified from the midbrain and striatum for bulk RNA sequencing. In the parkinsonian midbrain, CD11b(+) cells presented a unique anti-inflammatory transcriptomic profile that differed from degenerative microglia signatures described in experimental models for other neurodegenerative conditions. By contrast, striatal CD11b(+) cells showed a pro-inflammatory state and were similar to disease-associated microglia. In the midbrain, a prominent increase of infiltrated monocytes/macrophages was observed and, together with microglia, participated actively in the phagocytosis of dopaminergic neuronal bodies. Although striatal microglia presented a phagocytic transcriptomic profile, morphology and cell density was preserved and no active phagocytosis was detected. Interestingly, astrocytes presented a pro-inflammatory fingerprint in the midbrain and a low number of differentially displayed transcripts in the striatum. During alpha-synuclein-dependent degeneration, microglia and astrocytes experience context-dependent activation states with a different contribution to the inflammatory reaction. Our results point towards the relevance of selecting appropriate cell targets to design neuroprotective strategies aimed to modulate the innate immune system during the active phase of dopaminergic degeneration.

Despite the potential of CAR-T therapies for hematological malignancies, their efficacy in patients with relapse and refractory Acute Myeloid Leukemia has been limited. The aim of our study has been to develop and manufacture a CAR-T cell product that addresses some of the current limitations. We initially compared the phenotype of T cells from AML patients and healthy young and elderly controls. This analysis showed that T cells from AML patients displayed a predominantly effector phenotype, with increased expression of activation (CD69 and HLA-DR) and exhaustion markers (PD1 and LAG3), in contrast to the enriched memory phenotype observed in healthy donors. This differentiated and more exhausted phenotype was also observed, and corroborated by transcriptomic analyses, in CAR-T cells from AML patients engineered with an optimized CAR construct targeting CD33, resulting in a decreased in vivo antitumoral efficacy evaluated in xenograft AML models. To overcome some of these limitations we have combined CRISPR-based genome editing technologies with virus-free gene-transfer strategies using Sleeping Beauty transposons, to generate CAR-T cells depleted of HLA-I and TCR complexes (HLA-IKO/TCRKO CAR-T cells) for allogeneic approaches. Our optimized protocol allows one-step generation of edited CAR-T cells that show a similar phenotypic profile to non-edited CAR-T cells, with equivalent in vitro and in vivo antitumoral efficacy. Moreover, genomic analysis of edited CAR-T cells revealed a safe integration profile of the vector, with no preferences for specific genomic regions, with highly specific editing of the HLA-I and TCR, without significant off-target sites. Finally, the production of edited CAR-T cells at a larger scale allowed the generation and selection of enough HLA-IKO/TCRKO CAR-T cells that would be compatible with clinical applications. In summary, our results demonstrate that CAR-T cells from AML patients, although functional, present phenotypic and functional features that could compromise their antitumoral efficacy, compared to CAR-T cells from healthy donors. The combination of CRISPR technologies with transposon-based delivery strategies allows the generation of HLA-IKO/TCRKO CAR-T cells, compatible with allogeneic approaches, that would represent a promising option for AML treatment.

Understanding how genotypic variation results in phenotypic variation is especially difficult for collective behaviour because group phenotypes arise from complex interactions among group members. A genome-wide association study identified hundreds of genes associated with colony-level variation in honeybee aggression, many of which also showed strong signals of positive selection, but the influence of these 'colony aggression genes' on brain function was unknown. Here we use single-cell (sc) transcriptomics and gene regulatory network (GRN) analyses to test the hypothesis that genetic variation for colony aggression influences individual differences in brain gene expression and/or gene regulation. We compared soldiers, which respond to territorial intrusion with stinging attacks, and foragers, which do not. Colony environment showed stronger influences on soldier-forager differences in brain gene regulation compared with brain gene expression. GRN plasticity was strongly associated with colony aggression, with larger differences in GRN dynamics detected between soldiers and foragers from more aggressive relative to less aggressive colonies. The regulatory dynamics of subnetworks composed of genes associated with colony aggression genes were more strongly correlated with each other across different cell types and brain regions relative to other genes, especially in brain regions involved with olfaction and vision and multimodal sensory integration, which are known to mediate bee aggression. These results show how group genetics can shape a collective phenotype by modulating individual brain gene regulatory network architecture. The genetic basis of collective behaviour is complex. Single-cell transcriptomics of honeybee brains and gene regulatory network analysis showed differences in brain gene regulation and gene regulatory network plasticity between aggressive soldiers and non-aggressive foragers.

Early hematopoiesis is a continuous process in which hematopoietic stem and progenitor cells (HSPCs) gradually differentiate toward specific lineages. Aging and myeloid malignant transformation are characterized by changes in the composition and regulation of HSPCs. In this study, we used single-cell RNA sequencing (scRNA-seq) to characterize an enriched population of human HSPCs obtained from young and elderly healthy individuals. Based on their transcriptional profile, we identified changes in the proportions of progenitor compartments during aging, and differences in their functionality, as evidenced by gene set enrichment analysis. Trajectory inference revealed that altered gene expression dynamics accompanied cell differentiation, which could explain aging-associated changes in hematopoiesis. Next, we focused on key regulators of transcription by constructing gene regulatory networks (GRNs) and detected regulons that were specifically active in elderly individuals. Using previous findings in healthy cells as a reference, we analyzed scRNA-seq data obtained from patients with myelodysplastic syndrome (MDS) and detected specific alterations of the expression dynamics of genes involved in erythroid differentiation in all patients with MDS such as TRIB2. In addition, the comparison between transcriptional programs and GRNs regulating normal HSPCs and MDS HSPCs allowed identification of regulons that were specifically active in MDS cases such as SMAD1, HOXA6, POU2F2, and RUNX1 suggesting a role of these transcription factors (TFs) in the pathogenesis of the disease. In summary, we demonstrate that the combination of single-cell technologies with computational analysis tools enable the study of a variety of cellular mechanisms involved in complex biological systems such as early hematopoiesis and can be used to dissect perturbed differentiation trajectories associated with perturbations such as aging and malignant transformation. Furthermore, the identification of abnormal regulatory mechanisms associated with myeloid malignancies could be exploited for personalized therapeutic approaches in individual patients.

Identification of new markers associated with long-term efficacy in patients treated with CAR T cells is a current medical need, particularly in diseases such as multiple myeloma. In this study, we address the impact of CAR density on the functionality of BCMA CAR T cells. Functional and transcriptional studies demonstrate that CAR T cells with high expression of the CAR construct show an increased tonic signaling with up-regulation of exhaustion markers and increased in vitro cytotoxicity but a decrease in in vivo BM infiltration. Characterization of gene regulatory networks using scRNA-seq identified regulons associated to activation and exhaustion up-regulated in CARHigh T cells, providing mechanistic insights behind differential functionality of these cells. Last, we demonstrate that patients treated with CAR T cell products enriched in CARHigh T cells show a significantly worse clinical response in several hematological malignancies. In summary, our work demonstrates that CAR density plays an important role in CAR T activity with notable impact on clinical response.

Motivation An important step in the transcriptomic analysis of individual cells involves manually determining the cellular identities. To ease this labor-intensive annotation of cell-types, there has been a growing interest in automated cell annotation, which can be achieved by training classification algorithms on previously annotated datasets. Existing pipelines employ dataset integration methods to remove potential batch effects between source (annotated) and target (unannotated) datasets. However, the integration and classification steps are usually independent of each other and performed by different tools. We propose JIND (joint integration and discrimination for automated single-cell annotation), a neural-network-based framework for automated cell-type identification that performs integration in a space suitably chosen to facilitate cell classification. To account for batch effects, JIND performs a novel asymmetric alignment in which unseen cells are mapped onto the previously learned latent space, avoiding the need of retraining the classification model for new datasets. JIND also learns cell-type-specific confidence thresholds to identify cells that cannot be reliably classified. Results We show on several batched datasets that the joint approach to integration and classification of JIND outperforms in accuracy existing pipelines, and a smaller fraction of cells is rejected as unlabeled as a result of the cell-specific confidence thresholds. Moreover, we investigate cells misclassified by JIND and provide evidence suggesting that they could be due to outliers in the annotated datasets or errors in the original approach used for annotation of the target batch.

Single-cell RNA-Sequencing has the potential to provide deep biological insights by revealing complex regulatory interactions across diverse cell phenotypes at single-cell resolution. However, current single-cell gene regulatory network inference methods produce a single regulatory network per input dataset, limiting their capability to uncover complex regulatory relationships across related cell phenotypes. We present SimiC, a single-cell gene regulatory inference framework that overcomes this limitation by jointly inferring distinct, but related, gene regulatory dynamics per phenotype. We show that SimiC uncovers key regulatory dynamics missed by previously proposed methods across a range of systems, both model and non-model alike. In particular, SimiC was able to uncover CAR T cell dynamics after tumor recognition and key regulatory patterns on a regenerating liver, and was able to implicate glial cells in the generation of distinct behavioral states in honeybees. SimiC hence establishes a new approach to quantitating regulatory architectures between distinct cellular phenotypes, with far-reaching implications for systems biology.