Bioinformatics-Compute-Server

我们开发创新的生物信息学方法，用于现代分子生物学，生物医学和系统/网络医学。我们需要计算硬件来存储和分析大量所谓的组学数据。我们的目标是提取新的知识，并产生新的假设，如机制生物标志物。我们采用先进的投标数据技术以及人工智能和机器学习。我们的研究重点在于多尺度组学数据（如转录组学和表观基因组学）在时空分辨率与分子信号通路的组合分析。后者被建模为具有节点和它们之间的边的图。我们为高通量组学数据（TB级的序列和光谱数据）和基因组规模的网络（具有数千个节点和数十万个边缘）开发创新的启发式方法，以提供系统（生物）医学结果和目标。这些后来被用作机制生物标志物，并可能有助于，例如在医学中，识别新的药物靶点，建议药物再利用候选人，并找到多种药物的协同机制（网络药理学）。这需要一个本地高度并行的计算服务器基础设施，具有许多处理器核心（>48）和大内存（>256 GB），可以由所有计算核心并发访问。例如，我们的莱布尼茨研究中心提供的典型服务并不涵盖这一特定领域。然而，这样的系统是在5-10个时间点和5-10个位置将数百个生物医学样品中的数万个基因和小RNA彼此关联并与分子信号传导网络关联以有效地识别差异改变的热点的唯一方法。这些斑点是我们旨在产生的新颖的机制假设，可以测试其药物性并在之后的湿实验室中验证。

We develop innovative bioinformatics approaches for applications in modern molecular biology, biomedicine and systems/network medicine. We require compute hardware allowing for storing and analyzing massive so-called omics data. Our aim is to extract new knowledge and to generate novel hypothesis, such as mechanistic biomarkers. We employ advanced bid data technologies as well as artificial intelligence and machine learning. Our research focus lies on the combined analysis of multi-scale omics data (e.g. transcriptomics and epigenomics) in spatiotemporal resolution together with molecular signaling pathways. The later are modeled as graphs with nodes and edges between them. We develop innovative heuristic methods for high-throughput omics data (terabytes of sequence and spectrometry data) and genome-scale networks (with thousands of nodes and hundreds of thousands of edges) to deliver systems (bio)medicine results and targets. Those are used afterwards as mechanistic biomarkers and may help, for instance in medicine, to identify novel drug targets, to suggest drug repurposing candidates, and to find synergistic mechanisms of multiple drugs (network pharmacology). This requires a local highly parallel compute server infrastructure with many processor cores (>48) and big main memory (>256 GB), which can be accessed concurrently by all compute cores. The typical services offered, for instance, by our Leibniz Rechenzentrum, do not cover this specific area. However, such a system is the only way to associate tens of thousands of genes and small RNAs over hundreds of biomedical samples over 5-10 time points and 5-10 locations with each other and with the molecular signaling networks to effectively identify differentially altered hot spots. Such spots are the novel, mechanistic hypotheses that we aim to generate, which can be tested for drugability and validated in wet labs afterwards.