The H3Africa reference panel is open-access through the AfriGen-D
Imputation Service. You do not need to apply separately for the panel —
creating an account on the service gives you access.

1. Go to https://fedimpute.afrigen-d.org (or the legacy mirror
   https://impute.h3africa.org).
2. Sign in with your AfriGen-D identity (Kibali SSO). New users will
   be prompted to register on first sign-in.
3. Start a new imputation job — the H3Africa panel will be available
   in the panel selector.

The reference panel itself is not distributed for local use. All
imputation runs against the panel must go through the service. If you
have a use case that requires offline access (e.g. running a local
Nextflow pipeline against your own reference panel), the team can
discuss alternatives via helpdesk.

The H3Africa reference panel is built on GRCh38. The imputation
workflow can take input data in either GRCh37 or GRCh38:

• If you select the GRCh37/hg19 build at upload, the workflow
  automatically performs a liftOver to GRCh38 to match the panel.
• If your data is already on GRCh38, select that build at upload —
  no liftover runs.

Build 37 reference panels have been discontinued, so output is always
on GRCh38.

If you have a large fraction of allele mismatches against the panel
(e.g. >100 allele switches after liftover), pre-process your data
with the Allele Switch Checker workflow before re-running imputation
— see the related KB article on checkref.

The encryption password is shown on the job detail page, near
the bottom, in a line that reads something like:

All results are encrypted with password ...

Copy that password and use it to unzip each chromosome file. Example:

unzip -P 'YOUR_PASSWORD' chr_22.zip
# or with 7-Zip:
7z e chr_22.zip -p'YOUR_PASSWORD'

If you set a custom password when submitting the job, use that one
instead — it overrides the auto-generated password.

If unzip reports incorrect password, the most common cause is
quoting: the password may contain shell-special characters, so
always wrap it in single quotes. Copy it exactly as displayed
(no leading or trailing whitespace).

Bulk download via curl | bash is intentionally blocked for security
reasons. Use wget per file instead. The job page lists a
separate URL for each chromosome's .zip — copy each URL and run:

wget 'https://fedimpute.afrigen-d.org/share/results/<hash>/chr_22.zip'

If wget reports 400 bad request on multiple chromosomes, you've
likely exceeded the download-attempt rate limit. Wait a few minutes
and retry, or resubmit the job with a fresh job name — the
auto-generated download URLs and password will reset.

There is a per-job cap on total download attempts; this protects
the shared infrastructure from accidental wget loops.

The current AfriGen-D imputation service offers three H3Africa
reference panels (all on GRCh38):

• H3Africa v6 (full) — 4,447 samples, 58.7M variants.
  Sub-Saharan African ancestry plus additional cohorts.
• H3Africa v6 African-only — 1,895 samples, 42.3M variants.
  Strictly African-ancestry subset of v6.
• H3Africa v7 — 6,213 samples, 78.6M variants. Newer release
  with an expanded sample set.

AGVP and 1000 Genomes are not directly included as a labelled
subset — the panels are H3Africa-derived. HapMap2 CEU (60 samples,
chr 20 only) is also available for testing.

A summary description and per-population allele frequency
information is shown on each panel's detail page in the Explorer.

An HLA-specific reference panel is in development. The team is
building a Nextflow-based HLA imputation pipeline using HIBAG with
validated African-specific reference panels; this will be integrated
into the AfriGen-D imputation service.

Until it ships, options are:

1. Submit a helpdesk ticket describing your use case — depending on
   timing, the team can run HLA imputation for you against the
   in-progress panel.
2. Use a publicly available HLA reference panel with HIBAG locally
   (the HIBAG documentation lists supported panels), accepting that
   non-African panels will have reduced accuracy for African samples.

There is no public download of the H3Africa HLA reference panel.

First, check the Logs tab on the job detail page — most pipeline
failures have a descriptive error in PIPELINE OUTPUT (red text).
Common causes:

• Allele mismatches against the reference panel — pre-process
  with the Allele Switch Checker workflow before retrying.
• Wrong genome build selected — the build dropdown must match
  your input data, not the panel.
• Chromosome naming — VCF chromosomes can be either 1, 2, …
  or chr1, chr2, …; either is accepted, but they must be
  consistent within a file.
• Empty / truncated VCF — verify with bcftools view -h your.vcf.gz | tail that the file has variant records.

If the Logs tab is empty or the error is not actionable, raise a
helpdesk ticket and include the Job ID (visible on the job
detail page). With the Job ID, the team can inspect the WES run
trace and identify the cause much faster than from a description
alone.

If your job completed but the output directories are empty, check
the QC Statistics card on the Overview tab — if quality control
filtered out all variants, the imputation step will produce no
output.

H3Africa array manifests are hosted at
https://chipinfo.h3abionet.org/downloads. Both v1 and v2 manifests
are available there, in the formats Illumina GenomeStudio expects.

The allele-checker tool, checkref, is on GitHub at
https://github.com/afrigen-d/checkref. It identifies and fixes
strand / allele mismatches between your input VCF and the imputation
reference panel.

You do not need to run it manually for routine imputation. The
Allele Switch Checker workflow on the imputation service runs
checkref for you and produces a corrected VCF as a downloadable
output. Use the standalone repository if you want to integrate the
check into your own Nextflow / local pipeline.

The imputation service uses single sign-on via Kibali (AfriGen-D's
identity provider). To sign up:

1. Go to https://fedimpute.afrigen-d.org and click Sign in with
   AfriGen-D Identity.
2. On the Kibali page, choose Don't have an account? Sign up,
   and fill in your name, email, username, and password.
3. After submitting, you'll receive an email asking you to verify
   your email address. Click the link in that email.
4. Once verified, you'll be redirected back into the imputation
   service and can submit jobs immediately. There is no manual
   activation step for H3Africa panel access — sign-up gives you
   access automatically.

If the sign-up form does not advance when you click submit:

• Make sure the password meets the policy (minimum 8 characters,
  mixed case + a digit).
• Open the browser console (F12) to check for JavaScript errors.
  If there's an error, screenshot it and raise a helpdesk ticket —
  this has been the cause of past account-creation failures.
• Try a different browser or clear site cookies for
  kibali.afrigen-d.org.