Near-unsupervised computational methods for exploring "omic" data

该项目旨在通过批判性地分析和比较现有方法，研究最近提出的有前途的机器学习范式“近无监督学习”。该项目还旨在开发更广泛的大数据分析新算法，并适应以下三种应用：宏基因组数据中的物种分离；建立将基因组信息与癌症药物敏感性联系起来的模型；在质谱代谢组学数据中鉴定不同的代谢物分布模式。潜在的结果包括增加对肥料对不同植物品种和新出现的植物疾病的有用性的了解，癌症的基因组变异以及对微生物的更多了解。

The project aims to investigate the recently proposed promising machine learning paradigm "Near Unsupervised Learning" by critically analysing and comparing existing methods. The project also aims to develop new algorithms in the broader spectrum of Big Data Analytics and their adaptation to the following three applications: species separation in metagenomic data; development of a model to relate genomic information to cancer drug sensitivity; and, the identification of distinct metabolite distribution patterns in mass spectrometry metabolomic data. The potential outcomes include increased understanding of the usefulness of fertilisers on different plant varieties and newly emerging plant diseases, genomic variations in cancer and more insights into microbes.