Designing a study that reads more than one mark
Reading two marks at once takes one of two shapes: both out of a single library, or both out of separate assays on the same samples. What each design shares decides what its results can be said to show.
01What this article answers
A multi-mark project is a single-mark project with a second measurement attached to it. The article works through what that attachment changes: which parts of the experiment the two results share, what a comparison between them is allowed to say, and where the interpretation stops.
The distinction that matters is between reading two marks out of one library and reading two marks out of two assays. In the first case the material and the chemistry are shared, so the two results come from the same molecules and the same conversion steps; in the second, each result carries its own workflow and the comparability of the two has to be argued rather than assumed.
02The distinction
The useful way to read any two-mark design is to ask what the two arms have in common, because that is exactly what can be compared without further argument.
| Design | What the two results share | What has to be argued | What it supports |
|---|---|---|---|
| Both marks from one library | The material, the library chemistry and the sequencing run | That the signal for each mark is attributed to the right mark, since both are read from one readout | A statement about how the two marks relate within the same molecules |
| Two assays on the same samples | Sample collection, handling and input quality | That the two workflows are comparable, including their own controls and what each can detect | A statement about how the two marks behave across the same sample set |
| One mark positional, one at transcript level | Sample handling and the input type | That the two results are not on one scale, because only one of them has a position | A hypothesis about a relationship, not a per-site measurement of both |
| A sequencing result with an orthogonal measurement | The material, and the RNA that is being interrogated | That the two methods answer different questions rather than the same one twice | An independent check on whether a mark is present, and at what level |
The line between the first two rows is the one that is most often blurred. Sharing a library is a real advantage for interpretation, but it is also a shared exposure: a step in the chemistry that biases one mark will bias the other in the same direction, and that will look like agreement.
The fourth row is where a second measurement earns its place. Position and composition are different properties of the same RNA, so a result from one of them is evidence in front of the other rather than a repeat of it.
03How to choose the design
- Write the sentence first. State the claim as a sentence about one mark and see whether it still needs the second one. If it does not, the study is a single-mark study with an extra assay attached.
- Decide what the second mark is doing. A control, a background and the other half of one claim each require a different design, and each has a different cost in samples and in interpretation.
- Keep one input type and one quality criterion across the arms. Both marks are measured on the same RNA, so any difference in how the material was handled is indistinguishable from a biological one.
- Compare within a mark before comparing across marks. The size and direction of a change are read against that mark’s own controls; across marks there is no common scale.
- Compare at one level. Site to site, or transcript to transcript. Mixing the two is allowed in a discussion, not in a figure that implies one measurement.
- Decide in advance what a flat second mark will mean. If a negative result in one arm invalidates the other, the design has made one claim out of two measurements.
04Limits and common misreadings
- Two results are not automatically two findings. A study is stronger when each mark can be interpreted on its own, which is a design property rather than a by-product of counting marks.
- Two enrichment profiles do not confirm each other. They are the same kind of measurement, so agreement between them repeats the method instead of testing it.
- A shared library constrains as much as it helps. Chemistry that biases one mark biases the other, so a shared step can produce a correlated artifact rather than a shared result.
- Ratios of different marks are not on one scale. A higher number for one mark is not more modification than for another; each ratio is read inside its own method.
- A flat context mark is a result. It does not have to move for the primary claim to hold, provided the design said in advance what it was there for.
06FAQ
Is one library for two marks always the better design?
It is the better design when the two marks are chemically related and are read through the same conversion, because the material, the library chemistry and the sequencing run are then all shared. It is not a general shortcut for reading any two marks, and a shared library is also a shared exposure.
Can the two marks be compared by their ratios?
Only within each mark. A ratio is produced by a specific method, so between methods there is no common scale to compare on, and a higher number for one mark is not more modification than for another. Each ratio is read inside its own method, against that method’s own controls.
What does orthogonal mean in this context?
A different kind of measurement on the same material — position against composition, for instance, so that a result from one is evidence in front of the other rather than a repeat of it. Two sequencing routes for the same property are not orthogonal to each other, because they share the assumptions of the readout.
Where should a two-mark project start?
From the sentence the study is meant to support, and from which mark that sentence is about. Write the claim first and see whether it still needs the second mark, then choose the second one for what it contributes, since a control, a background and the other half of one claim each need a different design.