SuperFold

Gregg Rice, 2014 — gmr@unc.edu

SHAPE-directed RNA secondary structure prediction using windowed partition function and MFE folding.

• See ARCHITECTURE.md for code structure.

---

Requirements

• Python 2.7
• RNAStructure
• matplotlib
• httplib2 (optional, for structure drawing)

RNAStructure Setup

Follow the instructions on the RNAStructure website for your platform.

Python Modules (matplotlib, httplib2)

For each module: download the source, extract, then install:

python setup.py install --user

---

Usage

python SuperFold.py RNA.map

All flags are optional. See --help for details:

python SuperFold.py --help

---

Input Files

.map — SHAPE Reactivity (required)

Generated automatically by the ShapeMapper pipeline. Columns: nucleotide #, SHAPE reactivity, error, sequence. T is converted to U automatically.

1   0.002512    0.053798    G
2   -0.034906   0.143529    T
3   -0.077852   0.257623    T

.mapd — Differential SHAPE (optional)

Columns: nucleotide #, differential SHAPE reactivity, std error, sequence, Z-factor.

1   -999.0  -999.0  G   -999.0
2   -0.0124 0.2673  U   -74.244
3    0.0951 0.0833  U   -2.349

Create a .mapd file with:

python differenceByWindowSHAPEMAP.py nmia.map 1m6.map nmia-1m6.mapd 25

Pass to SuperFold via --differentialFile.

ssConstraints.txt — Single-Strand Constraints (optional)

One nucleotide index per line:

34
35
36

ListofPKs_ds.txt — Pseudoknot Constraints (optional)

Paired nucleotide indices, one pair per line. Used to reassemble pseudoknotted positions in the final step:

34 78
35 77
36 76

DMS data (ShapeMapper >= 2.2.0)

If data was generated with --dms, run SuperFold with --DMS to use fold/partition commands compatible with DMS SM 2.2+ data.

---

Output

SuperFold creates folders automatically and appends a cryptographic hash to folder/file names to avoid collisions. A log file is written to the results folder.

Folder	Contents
partition/	Intermediate partition function calculations
fold/	Intermediate MFE fold calculations
results/	Merged partition function and MFE structures (merged.*)
regions/	Per-region structure files and plots

Arc plot key (base pair probabilities from partition function):

Color	Probability (%)
Green	> 80
Blue	> 30
Yellow	> 10
Gray	> 3

The ShannonSHAPE PDF shows Shannon entropy and SHAPE analysis. Region cut sites are written to the log file.

---

Troubleshooting

If no base pairs appear in the output, try a smaller window size:

--partitionWindowSize 1000
--foldWindowSize 1000

For window sizes under 1000, set --trimInterior 200 to include interior windows. Note that smaller window sizes bias toward shorter-range interactions.