About ampliconflow
pipeline, release contract, portal
16S amplicon studies are deposited as raw reads and described in prose. ampliconflow turns them into a versioned release: a count table, a taxonomy, per-run QC, checksums and a machine-readable manifest. This site is the reading surface for those releases, and every number on it is computed from the released tables at build time.
18
Studies
452
Samples
280.6k
Features
16.7M
Reads
From reads to a release
Each stage records what it did and with which tool version, so a release can be re-derived and a number on a page can be traced back to the file it came from.
-
01 Discover
find the runs and their metadata in the archive
-
02 QC
reads, quality, primers and identity, per run
-
03 Trim
cut primers and adapters off the ends
-
04 Denoise
errors and chimeras out, real variants kept
-
05 Cluster
variants grouped at 97% identity
-
06 Taxonomy
assign each variant against a reference
-
07 Table
one count matrix per method group
-
08 Publish
checksums, manifest, citable release
each stage is idempotent: a re-run continues from the last complete artifact, never from a half-written file
Where the data comes from, and where it goes
provenance
- 1 ENA / SRA / BioSamplethe authoritative record; this site links back to it and never replaces it
- 2 ampliconflowQC, trim, denoise, cluster, taxonomy, table, into a release
- 3 Releaseparquet + BIOM + FASTA + per-run QC + checksums + RO-Crate
- 4 Hugging Face datasetthe files, served with HTTP range reads
- 5 This sitestatic pages computed from the release at build time
how it is served
The eleven downstream modules
Each module states its own n and the test it used. A module that cannot run on a release is shown as a flagged gap, never as an empty frame or a zero.
Alpha diversity
observed, Chao1, Shannon, Simpson, evenness
Rarefaction
multinomial draws to a depth ladder
Beta diversity
Bray-Curtis, PCoA, UniFrac when a tree exists
Ordination
constrained RDA and CCA, trajectories
Statistics
PERMANOVA, PERMDISP, Mantel, distance decay, FDR
Network
signed co-occurrence, hubs, components
Community states
CLR k-means, batch-bias audit
Effects
effect sizes, variance partitioning, taxa screens
Phylogenetic
Faith's PD, weighted and unweighted UniFrac
Signal
Pagel's lambda, Blomberg's K
Spatial
Moran's I, variogram, gradient response
What this site does not claim
- It does not replace the archive; the ENA, SRA and BioSample records are linked, not copied.
- It does not present a pooled number as an experiment. Cross-study figures say that they are pooled.
- It does not hide gaps: missing QC, no tree, no coordinates and weak paper links are all labelled.
- It does not compare across method groups. Runs with different chemistry stay in separate tables.