RESUMEN
BACKGROUND: Long non-coding RNAs (lncRNAs) are pivotal players in cellular processes, and their unique cell-type specific expression patterns render them attractive biomarkers and therapeutic targets. Yet, the functional roles of most lncRNAs remain enigmatic. To address the need to identify new druggable lncRNAs, we developed a comprehensive approach integrating transcription factor binding data with other genetic features to generate a machine learning model, which we have called INFLAMeR (Identifying Novel Functional LncRNAs with Advanced Machine Learning Resources). METHODS: INFLAMeR was trained on high-throughput CRISPR interference (CRISPRi) screens across seven cell lines, and the algorithm was based on 71 genetic features. To validate the predictions, we selected candidate lncRNAs in the human K562 leukemia cell line and determined the impact of their knockdown (KD) on cell proliferation and chemotherapeutic drug response. We further performed transcriptomic analysis for candidate genes. Based on these findings, we assessed the lncRNA small nucleolar RNA host gene 6 (SNHG6) for its role in myeloid differentiation. Finally, we established a mouse K562 leukemia xenograft model to determine whether SNHG6 KD attenuates tumor growth in vivo. RESULTS: The INFLAMeR model successfully reconstituted CRISPRi screening data and predicted functional lncRNAs that were previously overlooked. Intensive cell-based and transcriptomic validation of nearly fifty genes in K562 revealed cell type-specific functionality for 85% of the predicted lncRNAs. In this respect, our cell-based and transcriptomic analyses predicted a role for SNHG6 in hematopoiesis and leukemia. Consistent with its predicted role in hematopoietic differentiation, SNHG6 transcription is regulated by hematopoiesis-associated transcription factors. SNHG6 KD reduced the proliferation of leukemia cells and sensitized them to differentiation. Treatment of K562 leukemic cells with hemin and PMA, respectively, demonstrated that SNHG6 inhibits red blood cell differentiation but strongly promotes megakaryocyte differentiation. Using a xenograft mouse model, we demonstrate that SNHG6 KD attenuated tumor growth in vivo. CONCLUSIONS: Our approach not only improved the identification and characterization of functional lncRNAs through genomic approaches in a cell type-specific manner, but also identified new lncRNAs with roles in hematopoiesis and leukemia. Such approaches can be readily applied to identify novel targets for precision medicine.
Asunto(s)
Leucemia , ARN Largo no Codificante , Animales , Humanos , Ratones , Diferenciación Celular/genética , Línea Celular Tumoral , Proliferación Celular/genética , Regulación Neoplásica de la Expresión Génica , Genómica , Leucemia/genética , ARN Largo no Codificante/genética , ARN Largo no Codificante/metabolismoRESUMEN
OBJECTIVE: Deep-learning techniques, particularly the Transformer model, have shown great potential in enhancing the prediction performance of longitudinal health records. Previous methods focused on fixed-time risk prediction, however, time-to-event prediction is often more appropriate for clinical scenarios. Here, we present STRAFE, a generalizable survival analysis Transformer-based architecture for electronic health records. MATERIALS AND METHODS: The input for STRAFE is a sequence of visits with SNOMED-CT codes in OMOP-CDM format. A Transformer-based architecture was developed to calculate probabilities of the occurrence of the event in each of 48 months. Performance was evaluated using a real-world claims dataset of over 130 000 individuals with stage 3 chronic kidney disease (CKD). RESULTS: STRAFE showed improved mean absolute error (MAE) compared to other time-to-event algorithms in predicting the time to deterioration to stage 5 CKD. Additionally, STRAFE showed an improved area under the receiver operating curve compared to binary outcome algorithms. We show that STRAFE predictions can improve the positive predictive value of high-risk patients by 3-fold. Finally, we suggest a novel visualization approach to predictions on a per-patient basis. DISCUSSION: Time-to-event predictions are the most appropriate approach for clinical predictions. Our deep-learning algorithm outperformed not only other time-to-event prediction algorithms but also fixed-time algorithms, possibly due to its ability to train on censored data. We demonstrated possible clinical usage by identifying the highest-risk patients. CONCLUSIONS: The ability to accurately identify patients at high risk and prioritize their needs can result in improved health outcomes, reduced costs, and more efficient use of resources.
Asunto(s)
Insuficiencia Renal Crónica , Humanos , Algoritmos , Registros Electrónicos de Salud , Probabilidad , Systematized Nomenclature of MedicineRESUMEN
Type 2 diabetes mellitus (T2DM) is associated with poor outcome after stroke. Peripheral monocytes play a critical role in the secondary injury and recovery of damaged brain tissue after stroke, but the underlying mechanisms are largely unclear. To investigate transcriptome changes and molecular networks across monocyte subsets in response to T2DM and stroke, we performed single-cell RNA-sequencing (scRNAseq) from peripheral blood mononuclear cells and bulk RNA-sequencing from blood monocytes from four groups of adult mice, consisting of T2DM model db/db and normoglycemic control db/+ mice with or without ischemic stroke. Via scRNAseq we found that T2DM expands the monocyte population at the expense of lymphocytes, which was validated by flow cytometry. Among the monocytes, T2DM also disproportionally increased the inflammatory subsets with Ly6C+ and negative MHC class II expression (MO.6C+II-). Conversely, monocytes from control mice without stroke are enriched with steady-state classical monocyte subset of MO.6C+II+ but with the least percentage of MO.6C+II- subtype. Apart from enhancing inflammation and coagulation, enrichment analysis from both scRNAseq and bulk RNAseq revealed that T2DM specifically suppressed type-1 and type-2 interferon signaling pathways crucial for antigen presentation and the induction of ischemia tolerance. Preconditioning by lipopolysaccharide conferred neuroprotection against ischemic brain injury in db/+ but not in db/db mice and coincided with a lesser induction of brain Interferon-regulatory-factor-3 in the brains of the latter mice. Our results suggest that the increased diversity and altered transcriptome in the monocytes of T2DM mice underlie the worse stroke outcome by exacerbating secondary injury and potentiating stroke-induced immunosuppression. Significance Statement: The mechanisms involved in the detrimental diabetic effect on stroke are largely unclear. We show here, for the first time, that peripheral monocytes have disproportionally altered the subsets and changed transcriptome under diabetes and/or stroke conditions. Moreover, genes in the IFN-related signaling pathways are suppressed in the diabetic monocytes, which underscores the immunosuppression and impaired ischemic tolerance under the T2DM condition. Our data raise a possibility that malfunctioned monocytes may systemically and focally affect the host, leading to the poor outcome of diabetes in the setting of stroke. The results yield important clues to molecular mechanisms involved in the detrimental diabetic effect on stroke outcome.