Research

Nanopore sequencing is a commonly used technology to sequence biological molecules such as DNA, RNA, and proteins. Translating their initial raw data, electrical signals, into human-readable sequences of characters (e.g., DNA characters of A, C, T, G), a process called basecalling, is costly and ineffective. We design solutions that can directly analyze nanopore electrical signals without basecalling. These solutions help us build solutions that better utilize resource constrained devices (e.g., mobile devices or drones) to enable in-the-field and real-time biological data analysis. Additionally, we explore integrating our signal analysis solutions into standard genomics pipelines further to improve their accuracy and speed.

Analyzing genomic data is challenging as solutions must analyze very large volumes of data quickly and accurately. Such an analysis usually requires designing effective algorithmic and machine learning solutions in the genome analysis pipeline (e.g., read mapping, de novo genome assembly, error correction, metagenomics, and basecalling). We explore improving the accuracy and speed of analyzing genomic data to better generate insights from them.

To substantially improve speed and energy-efficiency of the computational approaches in genomics, we explore hardware acceleration of the underlying analysis. To this end, we explore designing solutions for GPUs, FPGAs, as well as emerging technologies such as processing in-/near-memory (i.e., data-centric computing), analog computing, and neuromorphic computing.