Buying bioinformatics services without buying a report you cannot reproduce: what a bioinformatics service should hand over beyond figures, how targeted sequencing, targeted ngs, targeted next generation sequencing, an exome panel, complete exome sequencing and exome sequencing analysis differ in what the coverage can support, what next generation sequencing services quote for once an ngs sequencer, the ngs kits it runs and the analysis are separated, where ngs library preparation kits, an ngs library prep kit, ngs library prep kits, a library preparation kit, library preparation kits, a library prep kit, next generation sequencing library prep and library preparation ngs decide the experiment before any sequencer is booked, why cdna library preparation, an rna to cdna kit, rna prep handling and small rna library prep need different handling entirely, and when automated library prep or automated ngs library preparation stops being an optimisation and becomes the only way to keep batches comparable

Specifying the work

1. Start from the question and the effect size. Detection of a rare variant, quantification of a modest expression change and discovery of a novel transcript need different coverage, different replication and sometimes different chemistry. The statistical requirement sets the design, and it should be agreed before samples are prepared.
2. Choose the library chemistry against the molecule. Short fragmented inputs, degraded material, low input samples and small RNA each need their own preparation. A kit chosen for convenience rather than for the input type is the most common reason a run produces libraries that do not represent the sample.
3. Randomise across batches deliberately. Preparation batch, index set and flow cell lane all leave signatures. Distribute experimental groups across batches rather than processing one group at a time, and record the batch structure so it can be modelled later.
4. Agree the deliverables in writing. Raw reads, alignment files, the processing scripts with versions and parameters, the reference used, quality metrics and the intermediate outputs. A report with figures and no underlying data is not a deliverable, it is a summary.
5. Confirm data retention and transfer before starting. Where the data lives, for how long, who pays for storage after the project and how it is delivered. Large data sets arriving without a transfer plan sit on a provider's server until the invoice for storage arrives.

Batch structure is part of the result

Samples prepared together resemble each other. If one experimental group was prepared on Tuesday and the other on Thursday, the difference you measure includes the day. This is not a subtle effect, and it is not removable by analysis once the design has confounded it.

Plan the batches at the same time as the groups, record which sample went into which batch with which index, and hand that table to whoever analyses the data. It costs nothing at the start and cannot be recovered at the end.

Reproducibility is a deliverable, not a virtue

An analysis that cannot be rerun is an analysis that cannot be corrected, extended or defended. The practical test is whether a competent person with the raw data and the handed over materials can regenerate a named figure without contacting the original analyst.

Write that test into the contract. It is the single clearest way to distinguish providers who hand over work from providers who hand over pictures.

Common questions

What should a bioinformatics deliverable include?

Raw and aligned data, the exact pipeline with software versions and parameters, the reference genome and annotation used, quality control metrics for every sample, and the code that produced each figure. Anything less cannot be reproduced or extended.

Is a targeted panel better than whole exome or genome?

Where the regions of interest are known, a panel gives deeper coverage for the same cost and a smaller analysis burden. Where they are not, broader sequencing avoids the ascertainment bias of having chosen the targets in advance.

Does library preparation need automation?

Above a few dozen samples, automation reduces the batch to batch variation that manual preparation introduces and makes the batch structure predictable. Below that, careful manual work with deliberate randomisation is usually sufficient.

Who should own the analysis in the long run?

Whoever will be asked to defend it. Outsourced analysis is efficient for well defined pipelines; exploratory work that will be iterated benefits from in house capability, or at least from a provider who hands over runnable code.

Browse by service class

• Compare Gene synthesis on advertised prices