What is RNA Profiling?

RNA Profiling identifies and presents major substructural trends in a Boltzmann sample. For shorter sequences, profiling also relates the major trends with each other in a summary profile graph, which further allows significant commonalities and differences to emerge.

How does profiling work?

Profiling works by identifying the most common helices found in the sample, both individually and in combination. In particular, equivalence classes are used to group together structures with trivial differences, and thresholds employed to eliminate low frequency elements. Employing both eliminates noise and results in highlighting major, high-frequency structural trends.

More specifically:

• Profiling first generates a Boltzmann sample of 1,000 structures through GTboltzmann.
• Next, it assigns every helix present in the sample to an equivalence class known as a helix class. Every time a helix is assigned to a helix class, the frequency of the helix class is augmented by 1.
• All the helix classes are then sorted in descending order of frequency.
• A threshold is determined, either by the program by default or by the user, and all helix classes above it are chosen as features.
• Each structure is 'profiled' according to its inclusion of features. Thus, every profile is a subset of features, and every profile is assigned to an equivalence class denoted by its common profile. Every time a structure is assigned to a profile, the frequency of the profile is augmented by 1.
• All profiles are sorted in descending order of frequency.
• A frequency threshold is determined, either by the program by default or by the user, and all profiles above it are chosen as selected profiles.
• Profiles have a partial order given by set inclusion, which can by visualized by a Hasse diagram known as the summary profile graph.

An example

Let's take a RNA sequence from the bacteria Vibrio cholerae involved in its quorum sensing pathway: VcQrr3. We stochastically sample 1000 structures from the Boltzmann ensemble of VcQrr3, which contains 82 unique helix classes. The default helix class threshold selects the top seven as features. These features are used to profile all the structures, which can be grouped into thirteen distinct profiles. The default profile threshold chooses the top four profiles as selected profiles, which are then related by the summary profile graph.

Reference

Emily Rogers and Christine E. Heitsch. "Profiling small RNA reveals multimodal substructural signals in a Boltzmann ensemble" Nucl. Acids Res. 2014 doi: 10.1093/nar/gku959