Experimental Design Best Practices

Talk to us early. We often cannot save a poorly designed experiment, but we can help you design a successful one. We are committed to helping you determine the optimal experimental design given your budget, including sample size, number of replicates, sources of batch effects, sequencing depth, and overall bioinformatics considerations for the experiment.

General Principles

Minimum 3 biological replicates per condition
Plan for confounders: batch effects, sex, passage number, time of collection
Include appropriate controls: input, IgG, vehicle-treated, matched normal
Discuss sample size and statistical power with CCBR before proceeding
Plan metadata collection early so study design, sample annotations, processed outputs, and raw files are ready for downstream analysis and GEO submission

≥3 biological replicates per condition (≥5–6 for primary patient tumors)
≥30–50M reads/sample; ≥100M for fusions or rare transcripts
Poly-A for RIN ≥7; ribo-depletion for degraded RNA
Never pool biological replicates
Discuss batch layout if samples span multiple sequencing runs

Use biological replicates (≥2 minimum, ≥3 preferred) and process all samples under identical conditions.
Use fresh or properly preserved samples with high viability and minimize mitochondrial contamination.
Avoid PCR over-amplification and monitor library complexity early.
Use paired-end sequencing and aim for approximately 50M paired-end reads per sample.
Distribute samples across batches and lanes to reduce confounding batch effects.

Use biological replicates (minimum 2, preferably 3 or more) under the same workflow conditions.
Use ChIP-grade, preferably ENCODE-validated antibodies and track lot numbers carefully.
Always include input DNA and matched controls; use spike-ins when comparing global changes.
Choose depth and read strategy based on target type, expected signal, and genome complexity.
Consider CUT&RUN or CUT&Tag when working with low-input or weak-signal samples.

Use biological replicates (minimum 2, preferably 3 or more) and process samples under identical conditions.
Use validated antibodies and consistent input material with healthy, gently handled cells.
Include IgG or no-antibody controls and maintain consistent spike-in ratios across samples.
Prefer paired-end sequencing and design library prep to capture the short fragments produced by CUT&RUN.
Balance samples across batches and avoid PCR over-amplification.

Matched tumor and germline pairs are ideal for somatic variant calling.
For mouse studies, matched tumor and germline samples from the same individual remain the preferred design; cohort-matched controls can be used with care.
Aim for ≥100× mean target depth for tumor samples and ≥50× for germline samples.
Always document the exome capture kit and retain the corresponding target BED file.
For familial studies, consult before sequencing because sample selection strongly affects downstream analysis.

Target ≥30× mean depth for germline analysis and ≥60–100× for somatic variant detection.
Always include a matched normal for somatic analyses whenever possible.
WGS is preferred when structural variation or copy-number variation is a major objective.
PCR-free library preparation is preferred for reducing GC bias and improving coverage uniformity.
For mouse studies, matched tumor and germline from the same individual remains the strongest design.

Use at least 2 biological replicates per condition, with 3 or more preferred when feasible.
Start with the highest-quality sample possible and aim for viability >80%.
A practical target is approximately 5,000–10,000 recovered cells per sample.
Start with a minimum of 20,000 read pairs per cell for gene-expression libraries.
Use multiplexing strategies when appropriate to reduce batch effects and improve throughput.

Select the appropriate workflow for the specimen type and perform a pilot study for new tissue types.
Optimize section thickness, tissue placement, and avoid damaged or low-quality regions.
Use independent biological replicates and distribute conditions across slides and batches.
Plan sequencing depth based on assay type, tissue complexity, and tissue-covered spots.
Capture metadata consistently and consider matched scRNA-seq or snRNA-seq reference data for downstream interpretation.

For further assistance in planning your NGS experiment, please contact nciccbr@mail.nih.gov or visit our Bethesda office hours on Thursdays from 3:00 to 5:00 PM. For cost information and assistance with experiment setup, please visit the Frederick Sequencing and Genomics Core (FSGC) website or contact Bao Tran at bao.tran@nih.gov.