We implemented a series of algorithmic optimizations to improve the performance of SEQMINER: first, SEQMINER supports directly reading/writing compressed and tabix‐indexed files. To support efficient random information retrieval from large data files, we incorporated and extended the tabix library into SEQMINER. Tabix proceeds by indexing blocks of compressed data files (bgzip) format. Using the binning index and linear index, the tabix library allows the quick location of the sequence data from disks that overlap the query interval. This design allows the retrieval of sequence information at a time complexity of O(log(N)). The original tabix library only allows storing all retrieved information as strings, which cannot be directly analyzed in R. In SEQMINER, we extended tabix and implemented features to randomly access files in METAL/RAREMETAL format. Retrieved information is automatically parsed, converted to the appropriate data types (as strings, floating numbers, etc.) and made available for analysis as standard R objects, e.g., list or data frames.
