MARTi consists of two main components:
- The MARTi Engine - a back-end which performs the analysis and can be a single desktop/laptop or a high performance compute (HPC) cluster.
- The MARTi GUI - a lightweight web-based front-end which allows users to view analysis results.
There are a number of possible ways to install MARTi, but in the following sections, we highlight the two most common:
- a local analysis configuration where both the MARTi Engine and the MARTi GUI are installed on a single laptop/desktop.
- an HPC configuration, where the MARTi Engine resides on an HPC and the MARTi GUI resides somewhere else - e.g. on a laptop/desktop.
In the local analysis configuration:
- A sequencing device (e.g. MinION Mk1C or GridION) generates basecalled reads.
- On a laptop or desktop computer, the MARTi Engine analyses the sequence data. It accesses the data either by mapping drives or rsync'ing the data over a network connection.
- On the same computer, the MARTi GUI provides a server which enables web browsers to access analysis. This could be a browser on the same computer, or any browser on the same network.
In the HPC configuration:
- A sequencing device (e.g. MinION Mk1C or GridION) generates basecalled reads.
- On the HPC system, the MARTi Engine analyses the sequence data. It accesses the data either by mapping drives or rsync'ing the data over a network connection.
- On a laptop or desktop, the MARTi GUI provides a server which enables web browsers to access analysis. This could be a browser on the same computer, or any browser on the same network.
The diagram above illustrates the processes that take place inside the MARTi Engine:
- Basecalled reads first pass through a prefilter which removes reads that don't meet a quality or length threshold.
- Remaining reads are batched into chunks for speedy analysis.
- By default, MARTi classifies reads with a combination of BLAST and its own Lowest Common Ancestor algorithm to assign reads to taxa. Centrifuge and Kraken2 can also be used as an alternative for lower powered systems.
- BLAST files are gzip compressed after analysis to reduce storage requirements.
- For AMR classification, MARTi uses BLAST and the CARD database.