---
title: AAV quickstart
description: Name a target tissue for per-serotype tropism reports with citations and no ranking, then codon-optimize a protein transgene against the 4.7 kb limit.
tags:
  - aav
  - quickstart
---

You are designing an AAV gene therapy and you need two things: a capsid that reaches your target tissue, and a transgene cassette that fits inside it. The [AAV tools](/docs/aav) do both in order — rank capsids for the tissue, then codon-optimize the transgene and check it against the packaging limit.

## Select a capsid

### Name the target tissue

Give the tissue you want to transduce: `cns`, `eye`, `heart`, `liver`, `lung` or `muscle`. Aliases such as `brain` or `retina` are accepted. This is the one required input.

### Narrow to one serotype, optionally

Set `serotype_preference` to get that serotype's full record back, and `max_payload_kb` to have your payload checked against the ~4.7 kb packaging limit. A serotype name that is not in the reference table is rejected with the list of known names, so a typo fails loudly rather than returning nothing.

There is no species, route-of-administration or neutralizing-antibody input. Species is reported per observation from the literature rather than requested.

### Read the reports, because nothing ranks them

Capsid selection runs as an asynchronous Modal job. It returns, per serotype, the tissues a cited study reports, the level that study reports, the species it was observed in, and the citation — in alphabetical order, with no score and no recommendation.

Two things to check before you shortlist anything: the `model` on each observation, since a `high` reported in mice is not a claim about humans; and whether a tissue is absent rather than low, because a tissue no source reports simply does not appear.

## Optimize the transgene

### Provide the transgene

Paste the **protein** sequence to optimize — an amino-acid sequence. The tool reverse-translates it, so the DNA is the output, not the input. A trailing `*` stop is stripped for you, and whitespace is ignored.

⛔ Do not paste DNA here. Nothing validates the input, and `A`, `C`, `G` and `T` are all valid amino-acid letters, so a nucleotide sequence is silently reverse-translated as if it were a four-residue protein. You get a successful job, a sequence roughly three times the length you sent, and a `protein_length_aa` equal to your DNA's base count. There is no error to warn you.

### Set the payload limit

Set the maximum payload size (default 4.7 kb). It is the only other input this tool takes, and it only decides whether a packaging verdict comes back.

Everything else about the run is fixed and not configurable: codon usage is optimized for *Homo sapiens* with no species parameter, GC content is held to 40–60% in a 60-nt window, translation fidelity is enforced, and `BsmBI`, `BsaI`, `EcoRI` and `BamHI` are avoided. There is no promoter, polyadenylation or regulatory-element input — the backend optimizes the coding sequence only and does not design a cassette, so assemble those around the sequence it returns.

### Submit and read feasibility

Transgene optimization runs as an asynchronous GPU job on Modal. The result card shows the optimized sequence, its length in kb, GC content, codon-optimization score, and whether it fits the AAV packaging limit.

## In chat

Describe the gene therapy and the model runs both steps in order.

**Try this prompt** — [open in a guest chat](/chat/new?guest=true&prompt=Design%20an%20AAV%20vector%20for%20retinal%20gene%20therapy%3A%20select%20capsid%20for%20photoreceptor%20tropism%20and%20optimize%20the%20RPE65%20transgene%20for%20codon%20usage%20and%20packaging%20size%20-%20does%20the%20optimized%20construct%20fit%20within%20the%204.7kb%20packaging%20limit%2C%20and%20provide%20a%20go%2Fno-go%20recommendation%20on%20the%20packaging%20feasibility%20before%20committing%20to%20a%20GMP%20manufacturing%20run%3F).

> Design an AAV vector for retinal gene therapy: select capsid for photoreceptor tropism and optimize the RPE65 transgene for codon usage and packaging size - does the optimized construct fit within the 4.7kb packaging limit, and provide a go/no-go recommendation on the packaging feasibility before committing to a GMP manufacturing run?

## Read next