Bridge Horizon
Users & use cases

Who Bridge Horizon is for

Four distinct roles, one verifiable record of how each result was produced.

Proof

A live tenant's run history

Real runs from a Horizon tenant. Each row links a submission to its pipeline, commit, speed tier, user, duration, and status. The same record is what reviewers see months later.

Pipelines visible
Bulk RNA-seq, scRNA-seq, QC
Commit recorded
rnaseq-50gb (2526a07)
Run durations
18m 34s — 45m 17s
Status surfaced
Completed / Failed
Bridge Horizon run history page: a filtered list of pipeline runs showing submission ID, pipeline name, commit, status, speed tier, user, created date, and duration.

Every run above operated on FASTQs in place. No file was downloaded, zipped, or re-uploaded to a separate analysis tool.

Who it's for

Four segments, one platform

01Core / NGS

Core Facility or NGS Service Provider

Deliver FASTQs and processed outputs with the pipeline that produced them, no shipped drives or shared S3 keys.

  • Customers analyze FASTQs in place — no download, zip, and re-upload to a separate analysis tool
  • Per-project or per-customer access, revocable with a set expiry date
  • Pinned pipeline version, reference, and container digest recorded with every FASTQ and BAM delivered
Pains addressed
08Transfer02Validation05Collaboration06Audit
›Detail — Returning sequencing output to PIs or paying customers

Files and the pipeline that made them, delivered together. Recipients re-run any step themselves.

  • One link replaces shipped drives and shared S3 keys
  • Every delivery carries its provenance, not a separate README
02Biotech

Small to Mid-size Biotech

Run RNA-seq, ATAC-seq, or variant-calling pipelines without command-line work.

  • nf-core/rnaseq (STAR, Salmon), nf-core/scrnaseq (Cell Ranger, STARsolo), nf-core/atacseq (BWA, MACS2), nf-core/sarek (BWA-MEM, GATK HaplotypeCaller / Mutect2)
  • BAM/CRAM, counts matrix, MultiQC HTML report, and VCF/gVCF returned with the exact run record
Pains addressed
01Friction02Validation04Iteration
›Detail — Running pipelines on your own samples without a dedicated bioinformatician

Submit analyses through a form. Outputs stay inspectable for later review.

  • No command line, no cluster account, no Docker setup
  • A bioinformatician can audit or extend any run later
03CRO

Preclinical CRO

Keep study metadata, sequencing runs, and analyses in one record per study.

  • Samples tagged with study arm, time point, and treatment group, joined to each run
  • Close-out export per study: FASTQs, BAMs, analysis notebooks, parameters, and access log
Pains addressed
05Collaboration06Audit07Continuity
›Detail — Delivering sponsor studies that link samples, runs, and analyses

One record per study links samples, runs, and analyses. Sponsors get a re-executable package at close-out.

  • No spreadsheet reconciliation between LIMS, sequencer, and analysis
  • Sponsor close-out is a single export, not a week of packaging
04Academic

Academic PI, Postdoc, or Graduate Student

Take a count matrix to figures without provisioning compute.

  • Start from a 10x count matrix, h5ad, or .rds without a local install
  • Seurat / Scanpy / DESeq2 / edgeR notebooks versioned alongside every figure
Pains addressed
01Friction04Iteration06Audit
›Detail — Working from a count matrix returned by a core facility

QC, clustering, and differential expression in the browser. Every figure links to the notebook that made it.

  • No local R or Python install, no HPC account needed
  • Figures stay reproducible months later, parameters intact
Pipelines & inputs

What you can run, on what

Named pipelines, version-pinned. The exact tool versions, container digests, and references are recorded with every run.

RNABulk RNA-seq›
Pipeline

nf-core/rnaseq

Tools
STARSalmonHISAT2RSEMfeatureCountsMultiQC
Inputs
.fastq.gz

+ sample sheet CSV

Outputs
.bam.tsv.html

gene counts, transcript quant, MultiQC report

SCSingle-cell RNA-seq›
Pipeline

nf-core/scrnaseq

Tools
Cell RangerSTARsoloAlevin-fryScanpy QC
Inputs
.fastq.gz10x outs/

+ sample sheet CSV

Outputs
.h5.mtx.h5ad.html

filtered feature matrix + web summary

GOPathway enrichment (from DGE)›
Pipeline

Bridge Informatics — Pathways Analysis

Tools
clusterProfilerReactomePABioconductor org.*

GO BP, KEGG ORA + GSEA; deterministic, no LLM

Inputs
DGE resultssample sheet CSV

per-row organism (human, mouse, rat, zebrafish, fly, yeast); multi-organism, multi-comparison runs in parallel

Outputs
ORA tablesGSEA rankedplots.html

GO BP / KEGG / Reactome, per-comparison report

DGEDifferential expression (from featureCounts)›
Pipeline

Bridge Informatics — DGE Pipeline

Tools
DESeq2featureCounts merge

deterministic, no LLM; MD5 checksums on outputs

Inputs
featureCounts .tsvsample sheet CSV

with pairwise comparisons

Outputs
count matrixDESeq2 results7 plots / comparison

normalized counts, log₂FC, padj; violin, volcano, MA, PCA 2D/3D, scree, heatmap

Additional nf-core pipelines (ampliseq, smrnaseq, hic, nanoseq) and custom workflows available on request. References for human and mouse pre-loaded; other organisms configurable at provisioning.

Why it exists

Why Horizon exists

Eight recurring pains in computational biology, and what Horizon does about each.

What it is not

Not a LIMS. Not a pipeline builder. Not a cloud notebook. Not a replacement for your bioinformatician.

01 / Friction

The infrastructure tax.

Before you run a single job, you're filing tickets. Waiting on IT. Explaining what a compute node is to someone who manages Windows laptops. Getting cloud infrastructure stood up is a weeks-long detour that has nothing to do with your science.

No servers to provision. Jobs run on managed AWS Batch; open a browser and submit.
02 / Validation

Building the pipeline before you can use it.

A gold-standard RNA-seq pipeline isn't something you download. It's assembled, containerized, stress-tested, and validated. A multi-week effort requiring real computational expertise. Most bench scientists either skip this step and accept substandard results, or wait for an informatician who's already buried.

Horizon ships named, version-pinned pipelines on day one: nf-core/rnaseq (STAR, Salmon), nf-core/scrnaseq (Cell Ranger, STARsolo), nf-core/atacseq (BWA, MACS2), nf-core/sarek (GATK HaplotypeCaller, Mutect2), nf-core/methylseq (Bismark). Each one tested against reference datasets before release.
03 / Engineering

Cloud integration is its own engineering project.

AWS Batch, IAM roles, S3 permissions, job queues. Connecting a bioinformatics pipeline to cloud compute is a discipline unto itself. It's not your discipline, and it shouldn't have to be.

Every pipeline runs on AWS Batch under the hood. You submit a job; Horizon handles the infrastructure, scaling, and teardown invisibly.
04 / Iteration

Re-running for every parameter tweak.

Adjusting a p-value threshold or fold-change cutoff should take seconds, like tasting a dish and adding salt. Instead it means filing a re-run request, joining a queue, and waiting hours or days. Want to change your sample-to-sample comparison? Same process, all over again.

Parameter changes are fully self-serve. Adjust thresholds, cutoffs, or sample groupings directly in the UI and resubmit in seconds. No ticket. No queue. No middleman.
05 / Collaboration

Sharing results is a multi-day production.

A completed analysis sitting on a server is worth nothing until your collaborators can see it. Exporting figures, packaging outputs, and transmitting securely can take days, during which your collaborator is blocked and the science is stalled.

A shareable link is generated when a run completes. Recipients see figures, outputs, and the exact parameter set without an export step.
06 / Audit

Reproducibility lives in someone's memory.

Eighteen months after a run, a reviewer asks exactly which pipeline version, parameter set, and input files produced Figure 3. Or a regulatory submission requires a complete computational audit trail. If that record wasn't captured automatically, reconstructing it is an archaeological dig, and sometimes the artifact is just gone.

Every Horizon run generates an immutable, timestamped log: inputs, parameters, software versions, outputs. Locked. Permanent. Ready for a publication methods section or a regulatory filing, on demand.
07 / Continuity

Data tied to a person, a server, or a contract.

Datasets stored on an individual's machine, a local server, or a CRO's infrastructure are one offboarding or budget cut away from disappearing. In a field where longitudinal data is irreplaceable, this is an unacceptable fragility.

Horizon provides year-long or perpetual data availability on managed cloud storage, independent of any individual, server, or third-party service relationship.
08 / Transfer

Download, zip, re-upload to the next tool.

FASTQs land on a core facility server. To analyze them in a separate tool, someone has to download tens of gigabytes, zip them, and re-upload to another platform. Multiply by every customer and every assay. Hours per delivery, terabytes of duplicate transfer, and a fresh chance for a corrupted file or a wrong sample sheet at each step.

FASTQs stay where they were produced. Pipelines run against them in place; recipients open the same files through the same record, with no shuttling between systems.