Services

AURORA

Automated Unified Reproducible Omics Reporting & Analysis

An NGS pipeline that takes raw FASTQ already on the server through QC, alignment, variant calling and expression quantification, and produces a report.

Open AURORA

What it does

AURORA (Automated Unified Reproducible Omics Reporting & Analysis) takes raw sequencing data and processes it so that the result does not change when the person does. It records every command it ran, without exception, and reads tool versions back out of the outputs, so months later you can retrace how those numbers were produced.

What to know before you start

The raw data must already be on the server. There is no FASTQ upload on this screen. You put the data in your own account folder under the designated data folder on the server, then pick that folder on screen. The only thing you can upload is a sample metadata CSV.

The reference genome is fixed to hg38. No other build and no other species can be selected right now.

Reference files and tools have to be installed on the server beforehand. Nothing is downloaded while a job runs. What is present and what is missing is shown as-is under 환경 점검 (Environment check).

Not everything you can see on screen actually runs. If a tool is not installed you can still select it, but the run will be blocked. After choosing a preset, always check 환경 점검 (Environment check) on the right. Which program each step uses and what is ready right now is set out in Software availability.

How you use it

The screen has four sections in the left sidebar — 새 분석 (New analysis) · 작업 현황 (Job status) · 결과 · 리포트 (Results and reports) · 관리자 (Admin) (admin only appears if you have the permission).

1. Source folder — create a folder with 프로젝트 만들기 (Create project) or pick an existing project. Drill down to where the raw data is. Samples are detected and shown as soon as you select it
2. Sample metadata — the detected samples appear as a table. Click a cell to edit it in place. You can also download it with CSV 내려받기 (Download CSV), fill it in, and send it back with CSV 올리기 (Upload CSV). If you fill in the comparison group (group), you can start a between-group comparison later on
3. Pipeline — choose the data type and a preset. Whatever the preset fixes is locked; you choose the rest step by step
4. Report details — client, institution, sequencer and so on. Blank fields are left out of the report
5. Run plan and environment check — the right-hand column shows what will run in what order and whether the required tools are installed, in real time
6. Start the analysis — if there is even one error the button will not respond
7. Job status — refreshes every 5 seconds. The analysis keeps running if you close the window. You can open a row, cancel a run, or delete it
8. Results and reports — seven tabs show the output

The seven result tabs

What is implemented

Data types accepted — WGS (whole genome) · WES (whole exome) · BarSEQ (cfDNA barcodes) · RNA-seq (transcriptome) · miRNA-seq (small RNA) · scRNA-seq (10x single cell).

Pipeline presets — GDC standard, GATK standard, somatic variants (BWA-MEM2 + Mutect2), BarSEQ (cfDNA), Tuxedo, STAR + featureCounts, miRSEQ, Cell Ranger. Only the ones that fit the data type are shown — for single cell there is only Cell Ranger. You can also pick no preset and specify everything yourself.

Tools you can choose per step

Command preview mode — records the commands that would run without actually running them. Use it to check your settings, or to get commands to run on another server.

Reproducibility record — every command run, tool versions (read back from the output files, and asked of the installed tool directly when they are not there) and 24 run settings go into the report appendix.

Reports — HTML and PDF. Eight sections, and they are produced even when the pipeline fails — the point is to show you how far it got.

Admin features — force a re-scan of the environment check, override tool and reference paths.

What it cannot do

You cannot upload FASTQ. Moving data onto the server has to happen outside this screen.

You can only work under your own account folder. Selecting the user folder itself as input is also blocked — you must pick a project folder beneath it.

Symlinked paths do not work. The path check resolves to the real path before comparing, so a folder set up as a link will not open.

There is no queue. Press 분석 시작 (Start analysis) and it runs immediately. Launch several and that many run at once — you have to manage resources yourself with 동시 실행 시료 (Concurrent samples) and the thread count.

There is nothing but hg38. Neither the reference build nor another species can be changed on screen.

Do not download large files through the browser. Downloads load the whole file into server memory before sending it, so for a large BAM you are better off taking it from the filesystem directly.

Switching to English leaves most of it in Korean. Table headers and cell values, buttons and notifications are fixed in Korean. The language switch only reaches the messages the server generates and some option names.

Some settings are not on screen. The Cufflinks reference GTF, the variant depth cutoff, the Cell Ranger core and memory settings and others run on defaults only. Changing them means an administrator has to put them in through a path override.

Cancelling a run is not complete. Only the runner and its direct children get the termination signal. Grandchild processes already running, such as STAR or BWA, can survive.

Rebuilding a report does not tell you the outcome. You only get a notification that the build started, and it stays quiet if it fails. Reopen the tab and check whether the file changed yourself.

Where people get stuck

It cannot find the FASTQ files. Check the filename convention. Read 1 has to look like .1.fastq · R1.fastq · _R1.fastq · _1.fq, and read 2 has to match it. 10x single cell requires the <sample>_S1_L001_R1_001.fastq.gz form — check that you pointed at the ..._10X_RawData_Outs folder the vendor gave you, as it is.

It says some samples have no mate. That is a warning, not an error. If the data is not single-end, check the filenames.

Trimming is off and alignment is blocked. BWA, TopHat2 and Bowtie2 read their input from the trimming output folder. Either turn trimming on or use STAR. STAR is the only exception.

You cannot turn off both quality check and trimming. The pipeline cannot build the input FASTQ list. At least one has to be on.

Somatic variants, but no pairs are found. Normal and tumour are told apart by a tag in the sample name. The defaults are ASN for normal and ASC for tumour. If the name does not contain those characters, no pair is made.

Exome without a BED file. Mutect2 searches the whole genome and the CNVkit reference gets noisy. Supply the capture BED.

Editing a sample ID causes an error. The sample ID becomes a folder name at every step. Use only letters, digits and . _ -. Duplicates and blanks block the run too.

A job is left as orphaned. The status file says it is running but there is no process. This happens when the server was restarted — delete it and start again.

The first request is slow. Reading the configuration takes about 7 seconds. Once per worker; after that it comes from cache. It is not downloading a reference genome.

Reading the output

Read QC summary → alignment rate → expression and variants, in that order. Which numbers you can trust is covered in Reading your results.

Software availability

AURORA does not do the computing itself. At every step it calls out to an outside program — BWA and STAR do the alignment, GATK does the variant calling. So if that program is not installed on the server, that step does not run.

환경 점검 (Environment check) on the AURORA screen tells you that before you run. This section is about how to read that screen, and what is ready right now.

The original paper for each program is collected in the References below.

Where to look on screen

The 새 분석 side is the one that matters. If even one of the required entries reads 없음 (missing), the 분석 시작 (Start analysis) button will not respond. It stops you before the start rather than halfway through — finding out the next tool is missing after hours of alignment is far worse.

What each step uses

The statuses below are as of 2026-08-24. The source of truth is always 환경 점검 on screen, so check there right before you run. When a tool is newly installed, that screen changes before this table does.

Note that SAMtools runs across several rows. Alignment (BWA), VarScan2 and Sequenza all call it alongside their own work. So one missing tool blocks three steps at once.

What can run end to end right now

Of the nine presets, only one runs right now: Tuxedo (RNA-seq). Trim Galore → TopHat2 → Cufflinks are all in place.

Put together yourself, STAR alignment + Cufflinks quantification works as well. Running only the quality check works too.

명령어 미리보기 (Command preview) is not subject to this. It is a mode that records the commands without actually running them, so it goes all the way through even with a tool missing. Use it to check your settings, or to take away commands to run on another server.

Reference files are checked too

Not just programs — reference files come up in the same table: the reference genome FASTA, the alignment indexes, VCFs such as dbSNP, COSMIC and gnomAD, the ANNOVAR and VEP databases, 18 of them in all.

This is what has been confirmed so far.

The exome capture BED is empty by default. Its absence does not block a run, but you have to supply it when you use Mutect2 or CNVkit on exome data — without it Mutect2 searches the whole genome and the CNVkit reference gets noisy. For the details see Preparing your data.

What to do when you are blocked

Installing is not something you can do. Programs and reference files have to be on the analysis server, and that server is the administrator's to touch.

1. In 환경 점검, note exactly which names read 없음
2. Pass those names to an administrator
3. If you are in a hurry, see whether another tool gets you around it — there is usually more than one option for the same job. If expression quantification is blocked, go to Cufflinks; if variant annotation is blocked, switch from ANNOVAR to VEP
4. If you do not need the results this minute, settle the settings with 명령어 미리보기 and run the same settings again once the tools are in place

For administrators: the actual install paths and how to override them are in the operations wiki. A program can be installed and still read 없음 — the place the configuration points at and the place it is actually installed are different, and the fix then is not a fresh install but 경로 덮어쓰기 (Path override) in the 관리자 tab.

Last updated · 2026-09-04