YUMIN HUANGPLANT GENOMICS
Home / Lab Docs

A working guide for the lab.

Shared defaults for starting well, moving projects forward, keeping analyses reproducible, and developing as an independent researcher.

Start here / 01

The first 30 days.

The first month should create a reliable working environment and a clear research question. Speed matters less than building habits that make later work trustworthy.

  • Set up institutional accounts, shared storage, computing access, Git, and a backed-up working directory.
  • Read three foundation papers and write one paragraph for each: the question, evidence, limitation, and next experiment.
  • Reproduce one published figure or summary table from raw or intermediate data.
  • Create a one-page project brief covering the biological question, study system, data, method, expected evidence, and main risk.
  • Agree on a 90-day deliverable and define what evidence would count as progress.

Project workflow / 02

Move from a question to evidence.

Projects should advance through explicit decisions. At each stage, record what was decided, why it was reasonable, and what result would change the plan.

01

Frame the question

State the biological claim, comparison, and measurable outcome in plain language.

02

Write the analysis plan

Specify samples, controls, inclusion rules, software, outputs, and decision points before scaling.

03

Run a small pilot

Use a representative subset to test assumptions, resources, file formats, and expected signal.

04

Scale with checkpoints

Automate repeated work and inspect quality summaries before launching each expensive stage.

05

Interpret alternatives

Ask which technical artifact or competing biological model could produce the same pattern.

06

Archive the evidence

Freeze inputs, scripts, environments, logs, and figure-ready tables before manuscript drafting.

Data & code / 03

A result must be traceable.

A colleague should be able to identify the inputs, rerun the analysis, and recover the plotted values without reconstructing decisions from memory.

  • Keep raw data read-only. Store checksums, source, date received, sample identifiers, and access conditions.
  • Use stable sample IDs in every file. Keep biological descriptions in a versioned sample sheet.
  • Put scripts and small configuration files under Git; do not commit raw sequencing data, credentials, or generated intermediates.
  • Record software versions, parameters, reference builds, random seeds, and computational environment.
  • Generate figures from tidy tables with scripts. Never use manual edits as the only record of a final figure.
project/
├── README.md          # question, owner, status, how to reproduce
├── metadata/           # sample sheets and data dictionary
├── config/             # parameters and environment files
├── workflow/           # scripts or workflow definitions
├── results/            # versioned summaries, not raw data
├── figures/            # scripted final figures
└── logs/               # run records and quality reports

Training / 04

A continuous path toward independence.

The milestones describe increasing ownership rather than fixed labels. Timing varies by background, project, and degree; progress is judged by evidence, judgment, and reproducibility.

Research apprentice

Reproduce a workflow, understand each major input and output, keep a usable record, and explain one figure with appropriate caution.

Junior researcher

Own an analysis module, compare methods, diagnose common failures, present evidence clearly, and propose the next discriminating test.

Independent researcher

Frame a tractable question, design the analysis, defend choices, integrate evidence, write a coherent manuscript, and help a newer member succeed.

Degree expectations

FIT TO STAGE

Master’s training should demonstrate sound analysis, reproducibility, and clear scientific communication within a focused scope. Doctoral training should additionally demonstrate question formation, methodological judgment, synthesis across evidence, and substantial intellectual ownership.

Communication / 05

Make the work easy to understand.

Good communication exposes uncertainty early, shortens debugging, and improves the scientific question. Bring evidence, not only a verbal summary.

  • Use a short weekly update: goal, completed work, evidence, blocker, and next decision.
  • For meetings, show the exact figure, command, error, or table behind the question.
  • When blocked, document two focused attempts, what each ruled out, and the smallest question that needs help.
  • Write figures before prose: one message per figure, with a title that states the result and a caption that states the evidence.
  • Discuss authorship and contribution expectations early, then revisit them as the work changes.

Reference SOP / 06

Bulk RNA-seq: alignment workflow.

This reference workflow covers standard gene-level quantification from paired-end reads. The project plan should define the reference build, strandedness, biological contrasts, and exclusion criteria before analysis begins.

Core sequence

WORKING REFERENCE
FASTQ + sample sheet fastp HISAT2 index HISAT2 alignment QC + counts

The minimum archive includes the sample sheet, reference FASTA and annotation identifiers, fastp reports, HISAT2 and samtools statistics, a gene-count matrix, software versions, parameters, logs, and the script that joins sample-level results.

  • Validate sample names, read pairs, checksums, and metadata before processing.
  • Run fastp with adapter detection and quality filtering; retain both HTML and JSON reports.
  • Build the HISAT2 index once from the declared reference; store the command, annotation splice sites, and reference checksum.
  • Align each sample with a consistent strandedness policy; sort and index BAM files and capture alignment statistics.
  • Quantify genes with a declared annotation and counting rule; verify library complexity, mapping rate, duplication, and sample relationships before differential analysis.