Why RNA Workflow Solutions Fail Reproducibility

Team Dynamic Matrices | 10 September 2026

A workflow that averages excellent yield and varies widely between runs is a bad workflow. That sounds obvious, and yet almost every product sheet for an RNA extraction, depletion or capture kit leads with a single yield or efficiency number and says nothing about spread. We think reproducibility, not yield, is the metric that decides whether a workflow belongs in a pipeline. Here is where variability comes from and how to audit for it.

Reproducibility failures get blamed on biology

When two runs disagree, the first suspects are the samples: a different patient, a different passage, a different day of collection. Sometimes that is right. Often the workflow itself is the source, and because nobody measured its run-to-run spread on a constant sample, the variance is invisible and gets attributed to biology. This is the same failure we describe for organoid matrices, and the fix is the same: measure the workflow on a fixed input before trusting it on variable inputs.

Three sources of run-to-run variability

1. Variable materials inside the kit

Many kits are built around a biologically sourced or manufacturing-tolerance-dependent component: an enzyme (RNase H, DNase, reverse transcriptase, proteinase K) with lot-variable specific activity; beads with lot-variable binding capacity; probe sets with batch differences in synthesis quality. Specifications state a minimum, not a value. Two lots can both meet spec and differ by 30 percent in activity, and that difference shows up as yield or depletion efficiency changing between the run you validated and the run you did in production.

Enzymes add a second variable: decay. Activity falls with time and with every freezer excursion, so a kit at month one and month eight are not the same kit.

2. Operator-dependent timing

Manual steps that depend on precise timing (a two-minute wash, a fast elution, a hybridisation that has to be stopped on time) introduce variability that compounds across operators and days. A step that is robust in the hands of the person who developed the protocol is often fragile in the hands of the fifth person to run it. Every "carefully" and "immediately" in a protocol is a variance source.

3. Scale amplifies everything

A workflow that varies by 10 percent between runs is tolerable at five samples a week. At 500 samples a week the same variance produces weekly troubleshooting, batch effects that need statistical correction, and occasional whole-plate failures. Small inconsistencies become structural once the workflow is a pipeline.

How to audit a workflow

Before adopting any kit or method into a pipeline, run it on a constant reference sample (a pooled RNA standard, or a commercial reference like the SEQC/MAQC samples) at least five times across at least two operators and two days. Report mean and coefficient of variation for yield, integrity and, for depletion or capture, efficiency. Then ask:

• Is the core chemistry fixed-specification or biological? Synthetic reagents manufactured to a defined composition behave the same every lot. Enzymes and extracts do not.
• Does published data show repeated runs, or one result? A single bar with no error bar is a demonstration, not a validation.
• Does the protocol rely on manual timing, or is consistency built into the chemistry? A capture that goes to completion and stays there is robust to a slow operator; one that has to be stopped at the right moment is not.
• Was it tested on your sample types? Performance on intact cell-line RNA says nothing about FFPE, plasma or organoid lysates.

What a fixed-specification chemistry changes

This is why we built LASSO without enzymes and without beads. The capture reagent is a synthetic polymer carrying synthetic DNA catcher strands, made to a defined specification; its composition and binding capacity are the same in every lot, and it is stable for about a year at 4 °C, so month one and month eight are the same reagent. Capture proceeds by hybridisation-triggered phase separation to completion and does not need to be stopped on time. The workflow has fewer manual transfers because there is no bead wash series. In our internal runs and in customer pilots the run-to-run coefficient of variation for rRNA depletion is in the low single digits, and we publish spread, not just means.

That does not make the rest of the workflow immune. Sample collection, lysis and library prep still need the same audit.