How to Learn Immunoinformatics: A Beginner’s Roadmap (2026)
Immunoinformatics uses computational tools to predict which fragments of a pathogen’s proteins the immune system recognises, then assembles the strongest fragments into a vaccine construct and tests that construct in silico. A complete beginner project runs from sequence retrieval through epitope prediction, construct assembly, docking and molecular dynamics, almost entirely on free web servers.
What is immunoinformatics, and what does a project actually involve?
Immunoinformatics is the computational side of immunology. Instead of screening peptides at the bench, you predict which short stretches of a pathogen protein (called epitopes) will be recognised by antibodies or by T cells, and you do it from sequence and structure alone.
The applied version most students meet is reverse vaccinology: start from a pathogen’s proteome, work down to a handful of predicted epitopes, join them into a single multi-epitope construct, and then show computationally that the construct is antigenic, non-toxic, non-allergenic, and capable of binding an innate immune receptor. The output is a candidate, not a proven vaccine. Every serious paper in this area ends with a call for laboratory validation, and yours should too.
This matters for how you present the work. A project that claims to have “designed a vaccine” overstates the result. A project that says it produced and computationally screened a candidate construct is defensible, and that is the framing reviewers accept.
Who this roadmap is for
BSc and MSc students in biotechnology, microbiology, bioinformatics or life sciences who can read a protein sequence and use a web browser. You do not need to program to complete a first project. Python helps later, when you want to batch predictions rather than paste sequences one at a time.
What software and servers do you need, and which are free?
Almost the entire workflow runs on free academic web servers. That is the practical reason immunoinformatics is a good first computational project: there is no licence to buy and, for most steps, nothing to install.
| Step | Tool | What it gives you | Cost |
|---|---|---|---|
| Sequence retrieval | UniProt, NCBI | Reference proteome and protein sequences | Free |
| Linear B-cell epitopes | BepiPred-3.0, ABCpred | Antibody-accessible stretches from sequence | Free |
| Conformational B-cell epitopes | ElliPro (IEDB) | Discontinuous epitopes from a 3D structure | Free |
| T-cell epitopes and MHC binding | NetMHCpan-4.1, IEDB Analysis Resource | Predicted binding to specific MHC alleles | Free |
| Toxicity screen | ToxinPred | Flags peptides predicted to be toxic | Free |
| Antigenicity and allergenicity | VaxiJen, AllerTOP | Screens candidate epitopes and the final construct | Free |
| Structure of the construct | AlphaFold DB, ColabFold | A 3D model to dock and simulate | Free |
| Receptor docking | AutoDock Vina, ClusPro, HADDOCK | Construct bound to TLR4 or another receptor | Free for academics |
| Stability check | GROMACS | Molecular dynamics of the complex | Free, open source |
| Expression planning | JCat | Codon-optimised sequence for a host | Free |
Two cautions on that table. First, the servers hosted at ddg-pharmfac.net (VaxiJen and AllerTOP) and the C-ImmSim immune simulator are academic servers that go offline for stretches, so plan for a tool being unavailable the week you need it and record which version you used. Second, “free for academics” is not the same as “free”: ClusPro and HADDOCK require you to register with an academic address and accept their terms.
What is the step-by-step reverse vaccinology workflow?
Nine steps, in the order a project actually runs. Each is a spoke guide in this series.
- Select the target proteins. Surface-exposed and secreted proteins from the reference proteome, because those are what antibodies can reach.
- Predict B-cell epitopes. Linear epitopes from sequence with BepiPred-3.0 and ABCpred; conformational epitopes with ElliPro if you have a structure.
- Predict T-cell epitopes. MHC class I and class II binders with NetMHCpan and the IEDB tools, against the alleles common in your target population.
- Filter hard. Keep only epitopes that are antigenic, non-allergenic, non-toxic, and conserved across strains. Most of your candidates die here, which is the point.
- Assemble the construct. Join surviving epitopes with linkers (AAY, GPGPG and EAAAK are the conventional choices) and add an adjuvant at the N terminus.
- Model the 3D structure of the assembled construct, then refine and validate it.
- Dock it to an immune receptor, usually TLR4 or TLR2, to show the construct can engage innate immunity.
- Run molecular dynamics on the complex to show the binding is stable rather than a docking artefact.
- Plan expression. Codon-optimise for your host and check the sequence would clone.
Steps 7 and 8 are where immunoinformatics meets the rest of computational structural biology, and where most students stall, because those steps need docking and simulation skills rather than web servers.
Steps 7 and 8 are the ones that stop most projects. Our live cohort covers structure preparation, docking and GROMACS molecular dynamics end to end, taught by researchers who have published in this field.
How long does it take to learn, and in what order?
Weeks 1 and 2: the biology and the epitope servers
Retrieve a proteome, choose targets, and run B-cell and T-cell prediction on one protein. Your goal is not a result, it is understanding what the scores mean and why a threshold is a choice you have to justify.
Weeks 3 and 4: filtering and construct design
Screen for antigenicity, allergenicity, toxicity and conservation. Assemble a construct with linkers and an adjuvant. By the end you should be able to explain why each epitope survived.
Weeks 5 to 8: structure, docking and dynamics
Model the construct, validate it, dock it to TLR4, and run a short molecular dynamics simulation. This is the part that transfers to any structural biology project, so it is worth learning properly rather than copying a protocol. Start with the docking roadmap and then the GROMACS roadmap.
Ongoing: writing it up
Record every tool version, every parameter and every threshold as you go. Reconstructing them at the end is where students lose weeks.
How does immunoinformatics relate to docking and molecular dynamics?
They are not competing fields. Immunoinformatics decides what to build, docking and molecular dynamics test whether the thing you built behaves.
A vaccine construct is just a protein, so once you have modelled it, the tools are the same ones used for any protein-ligand or protein-protein problem. That is why the docking and GROMACS guides in this series apply directly, and why an immunoinformatics project is a reasonable way into computational structural biology rather than a detour from it.
The reverse is also true. If you already know docking, adding epitope prediction upstream turns a docking exercise into a complete, publishable design study.
What mistakes make an immunoinformatics project unpublishable?
- Reporting predictions as findings. A predicted epitope is a hypothesis. Say so.
- Not recording tool versions. These servers update. A result from BepiPred-2.0 is not a result from BepiPred-3.0, and a reviewer will ask.
- Using default thresholds without justifying them. Every predictor has a cutoff that trades sensitivity against specificity. Choosing the default is fine, not knowing you chose it is not.
- Ignoring population coverage. MHC alleles differ by population. A construct optimised for alleles that are rare in your target population is a weak design, and the IEDB population coverage tool exists precisely for this.
- Skipping the conservation check. An epitope in a hypervariable region will not hold across strains.
- Treating docking as proof. A docking score alone does not establish binding. Molecular dynamics is what turns a pose into an argument.
What can you build for your CV or dissertation?
A complete multi-epitope construct against one pathogen, documented end to end, is a realistic single-semester MSc project and a genuine portfolio piece. It shows sequence handling, prediction, structural modelling, docking, simulation and scientific writing in one artefact.
To make it count, publish the workflow rather than just the conclusion: the target list, the thresholds, the surviving epitopes, the construct sequence, and the simulation settings. A reader who can reproduce your project is worth more to your application than a claimed result they cannot check.
Why trust these tools, and who maintains them?
The core resources are maintained by public research institutions rather than commercial vendors. The Immune Epitope Database and Analysis Resource is a curated public repository of experimentally characterised epitopes with a hosted suite of prediction tools. NetMHCpan and BepiPred are maintained by DTU Health Tech, and ToxinPred and ABCpred come from Prof. G. P. S. Raghava’s group. Each has an associated peer-reviewed paper you should cite alongside the tool.
On our side, immunoinformatics and vaccine design is the largest single topic in our team’s published work: nine peer-reviewed papers out of a research record that currently stands at 50 or more papers, cited over 950 times across 36 journals. The guides in this series are written against that experience rather than assembled from other tutorials.
Frequently asked questions
Do I need to know programming to start immunoinformatics?
No. A first project can be completed entirely through web servers. Python becomes useful when you want to run predictions across a whole proteome instead of pasting sequences one by one, and that is a reasonable second-project goal.
Is a computationally designed vaccine a real vaccine?
No. The output is a candidate construct supported by predictions. It has not been expressed, purified or tested in an animal model. Describing it as a validated vaccine is the single most common overstatement in student write-ups.
How long does a first immunoinformatics project take?
Roughly eight weeks of part-time work to go from proteome to a docked, simulated construct, assuming you are also learning docking and molecular dynamics along the way. The prediction steps are fast; the structural steps are what take time.
Which is better for B-cell epitopes, BepiPred or ABCpred?
They use different methods and disagree often, so the practical answer is to run both and treat the overlap as your higher-confidence set. Report both results rather than the one that suited your conclusion.
What free certification can I show for these skills?
We run free assessments with a verifiable certificate covering the underlying computational biology skills. See the free certifications hub.
Do these web servers ever go down?
Yes. Several are single-institution academic servers without commercial uptime guarantees. Note the date you ran each prediction, save the raw output, and do not leave a required server until the last week of your project.
Keep learning: related guides
- The computational biology skills roadmap, if you are deciding what to learn first.
- How to learn molecular docking, for step 7.
- How to learn molecular dynamics with GROMACS, for step 8.
- Our published research, including the immunoinformatics work behind these guides.
Want the structural half taught properly? The live cohort covers structure preparation, validation, docking and GROMACS molecular dynamics, with a working project at the end. Small group, taught by published researchers.