SHF: Small: Molecular Classifier Circuits for Disease Diagnostics

RNA和蛋白质分子水平的变化与大量人类疾病有关。监测这些变化使临床医生能够进行诊断、评估治疗效果和预测疾病复发。有时，仅检测一个分子标记就可以指示疾病状态，但更常见的情况是，有必要通过复杂的算法解释标记的组合，以获得可靠的诊断。在传统的诊断工作流程中，首先使用诸如RNA测序或微阵列之类的工具来检测和定量感兴趣的标记。然后使用计算机进行诊断，例如通过将测量结果与先前建立的基准进行比较。尽管这些方法在医学研究中被广泛使用，但对于需要定期监测或定期筛查的大量医疗应用来说，这些方法仍然成本高昂。为了克服这些限制，这项工作引入了一种新型的诊断工具，其中的计算和诊断是由一台“分子计算机”执行的，最大限度地减少了对复杂仪器的需要。这项研究与一个有两个主要目标的推广计划紧密结合在一起。第一个目标是开发一个教育项目，致力于教授在分子编程中取得成功所必需的跨学科技能。第二个和更长期的目标是增加工程研究中的女性入学人数。一个关键的目标是激励有电气工程和计算机科学背景的学生从事分子编程研究，方法是证明分子系统可以像我们对电子系统编程一样进行编程。为了实现这些目标，PI正在参与工程推广计划，并通过专业课程(例如合成生物学)和在核心电气工程和计算机科学课程中纳入分子编程模块，系统地将研究成果推向课堂。这项提议的目标是证明分子计算可以在疾病诊断中变得实用。该方法将计算机计算和试管计算有机地结合在一起。工作流程开始于对公开的基因表达数据进行计算分类器的训练-支持向量机(支持向量机)。然后，将In Silicon分类器映射到一组DNA链和复合体上，这些DNA链和复合体在分子水平上实现了相同的分类器，从而产生了一种新颖的分子计算体系结构。最后，在不同类型的分子数据上测试了该分子分类器。在初步工作中，Pi构建了一个分子支持向量机，原则上可以根据对七个宿主转录本的分析来区分细菌和病毒感染。当前提案的目标是优化和自动化分类器设计和测试，并使此类技术更接近实际应用。

Changes in the levels of RNA and protein molecules are associated with a large number of human diseases. Monitoring such changes enables clinicians to perform diagnosis, evaluate therapeutic efficacy and predict disease recurrence. Sometimes, detection of just a single molecular marker can be indicative of a disease state, but more commonly it is necessary to interpret a combination of markers via complex algorithms to obtain a reliable diagnosis. In a traditional diagnostic workflow, markers of interest are first detected and quantitated using tools such as RNA sequencing or microarrays. A computer is then used to make a diagnosis, for example by comparing the measurement results to a previously established benchmark. Despite their widespread use in medical research, these methods remain cost-prohibitive for a large number of medical applications where recurrent monitoring or regular screenings are necessary. To overcome these limitations, this work introduces a novel type of diagnostic tool where the computation and diagnosis is performed by a "molecular computer", minimizing the need for complex instrumentation.This research is tightly integrated with an outreach program that has two main goals. The first goal is to develop an educational program dedicated to teaching the interdisciplinary skills that are necessary to be successful in molecular programming. A second and longer term goal is to increase the enrollment of women in engineering research. A key aim is to motivate students with backgrounds in electrical engineering and computer science to engage in molecular programming research by demonstrating that molecular systems can be "programmed" just as we program electronic systems. To achieve these goals the PI is participating in engineering outreach programs and systematically pushes research results into the classroom, both through specialized classes (e.g. synthetic biology) and by incorporating molecular programming modules in core electrical engineering and computer science classes.The goal of this proposal is to demonstrate that molecular computation could become practically useful for disease diagnosis. The proposed approach integrates computation in silico with computation in the test tube. The workflow begins with the training of a computational classifier --- a support vector machine (SVM) --- on publicly available gene expression data. Then, the in silico classifier is mapped onto a set of DNA strands and complexes that realize the same classifier at the molecular level, resulting in a novel kind of molecular computation architecture. Finally, the molecular classifier is tested on different types of molecular data. In preliminary work, PI has constructed a molecular SVM that can, in principle, be used to distinguish between bacterial and viral infections based on analysis of seven host transcripts. The goal of the current proposal is to optimize and automate classifier design and testing and to bring such technology closer to practical applications.