Interdisciplinary model · INDSCI-SIM

ELiXSIR

Extended, zone-Linked
nine-compartment SIR

A COVID-19 simulation with nine disease compartments, age-stratified contacts, policy transitions, migration and Bayesian parameter estimation—used from district to national scales.

Domain: Epidemiology
Sampler: CosmoChord / MPI
Collaboration: INDSCI-SIM

Source on GitLab ↗

Model anatomy

ELiXSIR solves an extended SEIR system with nine compartments, age-dependent contact patterns and migration between zones. The framework allows different intervention scenarios to be studied. Migration may be coupled through a gravity prescription or a more generic connection structure.

The code was developed within the INDSCI-SIM collaboration, which provided scientific inputs during the Indian COVID-19 response, and integrates directly with samplers for parameter estimation. Its published application covered Karnataka districts and state, five major cities, and India as a whole—not Delhi alone.

Read the PLOS Computational Biology paper ↗

Figures from the source

Structure, movement and contact

ELiXSIR nine-compartment epidemic model and transition rates — **Nine-compartment anatomy**Disease states and transition rates used by the coupled differential-equation system.

Migration gravity matrix connecting districts in Tamil Nadu — **Migration coupling**A gravity-model coupling matrix linking regional populations through migration.

Age-stratified Chennai contact matrix during lockdown — **Contacts · lockdown**The age-structured contact matrix applied during a lockdown phase.

Age-stratified Chennai contact matrix during an unlocked phase — **Contacts · unlock**The corresponding contact structure after policy restrictions are relaxed.

Analysis pipeline

From priors to time-series bounds

Initialize

Set population, age groups, group fractions, compartment transitions and runtime constants.

Schedule

Use policy dates to switch contact matrices between lockdown and unlock regimes; model migration with a gravity or generic structure.

Infer

Supply priors, starting points, proposal widths and daily infection/death data; run CosmoChord samples through ELiXSIR over MPI.

Compare

Generate infections and deaths, evolve reporting bias, compare scaled infections and predicted deaths with reported series.

Reconstruct

Use GetDist for posterior distributions, then pass posterior samples back through ELiXSIR to obtain time-series bounds.

Published application

From districts to India

Karnataka districts

Serosurvey-calibrated estimates of infection and death undercounting.

Karnataka state

Joint fits to infections and deaths with uncertainty bands.

Five cities

Bengaluru, Chennai, Delhi, Mumbai and Pune, each with distinct epidemic structure.

India aggregate

A nationwide first-wave reconstruction with inferred infections and reporting bias.

Latent quantities

Inferred total infections, infection fatality ratio and the evolving reproduction number R(t).

Published ELiXSIR analysis of Karnataka COVID-19 cases, deaths and inferred epidemic quantities — Published result · PLOS Computational Biology · CC BY
One model, several observational layers
Reported cases and deaths are fitted jointly; posterior samples then recover inferred total infections, reporting bias, the infection fatality ratio and R(t).
**Read the paper and figure source ↗**