## RAG Sampler

RAG Sampler produces 3D tree graphs for a given RNA secondary (2D) structure and uses hierarchical graph-sampling to generate candidate graph topologies ^{[1, 2]}. These candidate graphs can be used as input to build atomic models as predicted 3D structures (see RAG Builder).

Please provide the **RNA 2D structure file** (.bpseq format). *This server currently does not accept RNA structures with over 200 nucleotides.*

For file format details, see Information page or download a sample 2D structure file. *To run RAG Sampler with the sample input, submit with the file field blank.*

## Show Algorithm and Options Description ▼

RAG Sampler uses our RAGTOP (RNA-As-Graphs Topology Prediction) hierarchical graph sampling protocol to generate candidate 3D graphs. The given 2D structure is converted into an initial 3D tree graph. Data Mining tools are then used to predict the coaxial stacking and family for helical arrangements for RNA junctions ^{[11, 12]}. Then, Monte Carlo/Simulated Annealing (MC/SA) sampling is executed at flexible internal loop vertices of the 3D tree graph. Each move is performed by randomly choosing an internal loop and one of its neighbouring helices and rotating it along a randomly selected axis (\(x\), \(y\), or \(z\)). For **random** moves (default), we use the full range (0 to 360 degrees). The range of the angle rotation for local or **restricted** moves is specified (e.g., 20 degrees). The **SA protocol** (default) consists of cooling the "system temperature" by the effective term $$T_i = {c \over {log_2(1 + {i \over s})}}$$ where \(c\) is a constant, \(i\) is the current MC move number, and \(s\) is the total number of MC moves specified *a priori* (50,000 steps). The orientation of the junction is fixed during the MC/SA simulation. A knowledge-based scoring function based off of existing RNA structures is used to score all of the sampled graph topologies.

See the full description of our MC/SA protocol: Graph-based sampling for approximating global helical topologies of RNA, Using sequence signatures and kink-turn motifs in knowledge-based statistical potentials for RNA structure prediction.

< Return to previous page