Institute of Information Science, Academia Sinica

Proliferation of information systems and technologies have created an enormous capability to generate huge amounts of data from all levels of scientific studies (e.g. biological experiments and drug compound development) or systems. This has resulted in accumulating large databases where a great part of it is in the form of time-series data. Many attempts have been made in the past to analyze and search for patterns in this class of data, perhaps the earliest being associated with forecasting and domains outside of life sciences.

Of the two major areas where we normally have access to large amounts of data: engineered systems (e.g. complex sensor based medical equipment) or systems designed by nature (e.g. life sciences), recently many forms of time-series data have become available in the latter (i.e. life sciences). This evolving paradigm has a direct influence on discovering useful patterns and models that can be validated for better applicability in the field.

This seminar will include a brief overview of time-series data, focusing on life science domain. We explain the motivations and challenges in analyzing certain class of short time-series data obtained from biological experiments involving high-throughput genomics or proteomics experiments. We present real-world case studies and demonstrate how our methodology can lead to some interesting discoveries in this field. The presented methodology can be applied in other domains as well.