Software and development quality in Vidjil

Quality relies both on software engineering methods (agile development, systematic testing, continuous integration and delivery) and on human and team processes. We plan to be as efficient as possible while keeping a friendly and open attitude.

This document details our current quality assurance policy. The Vidjil user manual and more technical information for the developers are found on Moreover, within the VidjilNet consortium, we offer additional support and documents for certification processes.

We develop Vidjil with systematic testing and continuous integration. As of 2020, we have 2700+ tests spanning all components of the platform. We always want to extend the test coverage and to improve our development policy. We aim to provide



We develop an algorithm to process high-throughput sequencing data to detect sequences with V(D)J recombinations as well as incomplete or uncommon recombinations. We gather these sequences into clonotypes. We use and improve state-of-the-art text algorithms (spaced seeds, automata, bit parallelism, indexing structures...) to provide efficient analytical methods. As far as possible, we do not compute full alignments. We always plan to improve the algorithm, improving accuracy and speed and providing more pertinent analysis for immunology and hematology. We benchmark and improve the algorithm on carefully curated sequences.

The original algorithm was published in [BMC Genomics 2014], and extension to several loci as well as incomplete recombinations have been described in [PLOS One 2016]. Vidjil-algo was evaluated as very reliable. As of 2020, we are still working in improving again the efficiency of the algorithm and use more efficient algorithms for seed patterns evaluation.


Each new feature or release is carefully tested and goes through the following tests:

Vidjil web application


We develop the web application (both client and server) to wrap or to link several software to pre-process, process, or post-process RepSeq data.

Developed in Javascript with jQuery and d3.js, the web client is made for the visualization, inspection and analysis of clonotypes and their tracking along the time in a MRD setup or in a immunological study. It visualizes data processed by vidjil-algo or any RepSeq clonotype gathering software as soon as they output a compliant JSON format (documented on This gives some modularity to users if they need to combine Vidjil-algo results with other data, coming from either personal an analysis or other software or scripts.

A sample database links the web application and the algorithmic part, allowing users to upload sequence files and manage their jobs directly from the web application. When uploading files, the user can choose some predefined preprocess to be launch on her data. A server, implemented in Python with the web2py framework (, queues the job requests, allowing many jobs to be scheduled without overloading the server.


Sofware engeneering methods

Continuous integration, continuous or qualified delivery

Continuous integration (CI) consists in systematically testing the modification brought to the code (either on the algorithm or on the web application). This is used since 2017, and the test coverage regularly improves.

Continuous delivery (CD) consists in systematically deploying new releases when the tests succeed. We aim at reaching this point on the public test server ( As of 2020, we have automated delivery that is still triggered by the core developpers.

However, on production servers used in routine analysis, we do not use CD but rather qualified delivery. In collaboration with members of the consortium, we identify releases to be deployed. Some consortium members have also they own qualification tests to double-check new releases, and we provide to them additional qualification documents, for example for certification purposes.

Tracking issues

Both bug reports and feature requests are followed on our public bugtracker,

Human and Team processes

User relations

Members of the VidjilNet consortium have access to the support with guaranteed response time.

Team organization

The Vidjil Team works remotely, in Lille, Rennes, and in other places in the world.

Interactions between the developers are frequent:

Developer relations

The Vidjil platform aims to interact with other RepSeq software. We both use and propose APIs to work within a ecosystem of RepSeq software. We contribute to open formats to exchange RepSeq data.

Maintenance and hosting

We maintain a public test server,, open to everyone and free to use. This server is only for test and should not be used for clinical purposes.

Throughout the VidjilNet consortium, we propose two options for production servers with clinical and patient data:

Backups and incidents

Incidents, either on our side, or due to the hosting operator, can always happen. However our goal is to ensure that no results or analyses could be lost. Since October 2014, we lost once 1.5 days of data analyzed by our users on the public server (0.5 day in 2016, 1.0 day in November 2017). No loss of data occurred in 2018 neither in 2019. We have improved our backup strategy to lower the likelihood of such an event.

On the public test server, the backup strategy is as follows: