RNA Binding Proteins (RBPs) Rotamer Libraries

A high-precision, circular-statistics-derived backbone-dependent (BBD) and backbone-independent (BBI) rotamer library suite for RNA-binding proteins in bound and unbound states across interface and surface locations.

---

📁 Directory Structure

RBPs_rotamer_lib/
├── RBPs_BBD_rotamer_lib/         # Backbone-Dependent (30° x 30° Phi/Psi bins) Libraries
│   ├── RBPs_I_bbd_B_rotamer_lib.[csv|tsv|json]   # Bound Interface (I)
│   ├── RBPs_I_bbd_U_rotamer_lib.[csv|tsv|json]   # Unbound Interface (I)
│   ├── RBPs_N_bbd_B_rotamer_lib.[csv|tsv|json]   # Bound Non-Interface Surface (N)
│   ├── RBPs_N_bbd_U_rotamer_lib.[csv|tsv|json]   # Unbound Non-Interface Surface (N)
│   ├── RBPs_bbd_B_rotamer_lib.[csv|tsv|json]     # Bound Overall Surface (B)
│   └── RBPs_bbd_U_rotamer_lib.[csv|tsv|json]     # Unbound Overall Surface (U)
├── RBPs_BBI_rotamer_lib/         # Backbone-Independent (Global) Libraries
│   ├── RBPs_I_bbi_B_rotamer_lib.[csv|tsv|json]   # Bound Interface (I)
│   ├── RBPs_I_bbi_U_rotamer_lib.[csv|tsv|json]   # Unbound Interface (I)
│   ├── RBPs_N_bbi_B_rotamer_lib.[csv|tsv|json]   # Bound Non-Interface Surface (N)
│   ├── RBPs_N_bbi_U_rotamer_lib.[csv|tsv|json]   # Unbound Non-Interface Surface (N)
│   ├── RBPs_bbi_B_rotamer_lib.[csv|tsv|json]     # Bound Overall Surface (B)
│   └── RBPs_bbi_U_rotamer_lib.[csv|tsv|json]     # Unbound Overall Surface (U)
├── input_files/                  # Cleaned input datasets (PRDBv3 187/288 complexes)
├── output_files/                 # Torsion angle dataset exports (CSV, TSV, JSON)
├── script/                       # Core python engines for derivation & analysis
└── utils/                        # Structure parsing utilities (mmcif_clean_reader)

---

⚙️ Implementation & Derivation Methodology

The rotamer library derivation pipeline incorporates modern circular statistics and kernel regression methods (Zhang et al., Bioinformatics 2024, doi:10.1093/bioinformatics/btae444):

1. Torsion Angle Calculation Engine

• Built on macromol_torsion (geometry.py, protein_backbone.py, protein_sidechain.py).
• Enforces strict IUPAC right-handed sign convention ($\pm 180^\circ$) matching BioPython calc_dihedral with 100.00% precision.
• Assigns 3-state secondary structure via DSSP (get_secondary_structure).

2. SASA & Spatial Residue Classification Logic

Residues are categorized based on surface accessibility and interface burial using PRince (.int and .sur outputs):

• Interface Residues ('I'): Residues present in PRince .int output files that lose solvent accessible surface area (SASA) upon complexation (buried in the binding interface).
• Non-Interface Surface Residues ('N'): Residues present in PRince .sur output files (surface-exposed residues) that are not involved in the binding interface.
• Core Buried Residues ('C'): Interior buried residues with zero surface accessibility (absent from both .int and .sur files).
• Surface Population ('S'): All surface-exposed residues combined ($\text{Surface } S = \text{Interface } I + \text{Non-Interface } N$, derived by filtering SASA.isin(["I", "N", "S"])).
• Overall Population ('ALL'): Complete dataset combining Interface, Non-Interface Surface, and Core Interior residues ($I + N + C$).

3. Depth-Valid Sample Masking & Symmetry Folding

• Multi-Chi Depth Validation: Residues with missing terminal atoms are validated up to their deepest non-null chi depth, ensuring partial sidechains contribute accurately without introducing NaN binning artifacts.
• Symmetric Planar Ring Folding: Symmetric terminal chi angles ($\text{ASP }\chi_2$, $\text{GLU }\chi_3$, $\text{PHE }\chi_2$, $\text{TYR }\chi_2$) are folded into $[ -90^\circ, 90^\circ )$ prior to rotamer binning to avoid splitting chemically identical conformations into separate states.

4. Von Mises Kernel-Weighted Circular Statistics

• Grid Binning: Backbone $(\Phi, \Psi)$ angles are binned into $30^\circ \times 30^\circ$ grid cells centered at $[-180^\circ, -150^\circ, \dots, +150^\circ]$.
• Rotamer States: Sidechain $\chi$ angles are binned into standard staggered states ($p / g^+$ at $+60^\circ$, $t$ at $180^\circ$, $m / g^-$ at $-60^\circ$).
• Circular Statistics: Weighted circular mean ($\mu = \text{angle}(\sum w_i e^{j\theta_i})$) and circular standard deviation ($\sigma$) are calculated using scipy.stats.vonmises fitting, eliminating artificial $-180^\circ / +180^\circ$ boundary discontinuities.

5. Unified Multi-Format Library Outputs

• All rotamer libraries are exported in unified CSV, TSV, and JSON files combining all 18 standard sidechain amino acids for each state and surface classification.

---

🌀 Backbone-Dependent (BBD) Rotamer Libraries

File Descriptions

File Name	Description
RBPs_I_bbd_B_rotamer_lib.[csv|tsv|json]	Bound interface rotamers ($I$, Bound $B$)
RBPs_I_bbd_U_rotamer_lib.[csv|tsv|json]	Unbound interface rotamers ($I$, Unbound $U$)
RBPs_N_bbd_B_rotamer_lib.[csv|tsv|json]	Bound non-interface rotamers ($N$, Bound $B$)
RBPs_N_bbd_U_rotamer_lib.[csv|tsv|json]	Unbound non-interface rotamers ($N$, Unbound $U$)
RBPs_S_bbd_B_rotamer_lib.[csv|tsv|json]	Bound overall surface rotamers ($S$, Bound $B$)
RBPs_S_bbd_U_rotamer_lib.[csv|tsv|json]	Unbound overall surface rotamers ($S$, Unbound $U$)
RBPs_bbd_B_rotamer_lib.[csv|tsv|json]	Overall surface bound rotamers (Bound $B$)
RBPs_bbd_U_rotamer_lib.[csv|tsv|json]	Overall surface unbound rotamers (Unbound $U$)

Column Schema

Column	Description
AA	Amino acid 3-letter code (e.g. ARG, LYS)
PHI	Backbone $\Phi$ bin center angle (degrees)
PSI	Backbone $\Psi$ bin center angle (degrees)
_Count	Number of observations in specific $(\Phi, \Psi, \chi)$ rotamer bin
Count	Total observations in the $(\Phi, \Psi)$ backbone bin
Prob	Conditional probability $P(\text{rotamer} \mid \Phi, \Psi)$
B_CHI[1-4] / U_CHI[1-4]	Weighted circular mean $\chi_n$ angle (degrees)
B_CHI[1-4]Sig / U_CHI[1-4]Sig	Weighted circular standard deviation of $\chi_n$ angle (degrees)

---

📊 Backbone-Independent (BBI) Rotamer Libraries

File Descriptions

File Name	Description
RBPs_I_bbi_B_rotamer_lib.[csv|tsv|json]	Bound interface rotamers ($I$, Backbone Independent)
RBPs_I_bbi_U_rotamer_lib.[csv|tsv|json]	Unbound interface rotamers ($I$, Backbone Independent)
RBPs_N_bbi_B_rotamer_lib.[csv|tsv|json]	Bound non-interface rotamers ($N$, Backbone Independent)
RBPs_N_bbi_U_rotamer_lib.[csv|tsv|json]	Unbound non-interface rotamers ($N$, Backbone Independent)
RBPs_S_bbi_B_rotamer_lib.[csv|tsv|json]	Bound overall surface rotamers ($S$, Backbone Independent)
RBPs_S_bbi_U_rotamer_lib.[csv|tsv|json]	Unbound overall surface rotamers ($S$, Backbone Independent)
RBPs_bbi_B_rotamer_lib.[csv|tsv|json]	Overall bound residues (Backbone Independent)
RBPs_bbi_U_rotamer_lib.[csv|tsv|json]	Overall unbound residues (Backbone Independent)

Column Schema

Column	Description
AA	Amino acid 3-letter code
_Count	Observations in specific rotamer state
Count	Total observations for amino acid
Prob	Global rotamer probability $P(\text{rotamer})$
CHI[1-4]	Global weighted circular mean $\chi_n$ angle (degrees)
CHI[1-4]Sig	Global weighted circular standard deviation of $\chi_n$ angle (degrees)

---

💻 Usage Example

import pandas as pd

# Load Backbone-Dependent Bound Interface Rotamer Library
df_bbd = pd.read_csv("RBPs_BBD_rotamer_lib/RBPs_I_bbd_B_rotamer_lib.csv")

# Filter rotamers for Arg at phi = -60, psi = -30
arg_rotamers = df_bbd[(df_bbd["AA"] == "ARG") & (df_bbd["PHI"] == -60) & (df_bbd["PSI"] == -30)]
print(arg_rotamers[["AA", "PHI", "PSI", "Prob", "B_CHI1", "B_CHI2", "B_CHI1Sig"]])

---

🔗 References

1. Kant, S., Nithin, C., Mukherjee, S., Maity, A., Bahadur, R.P., 2025. Protein–RNA Docking Benchmark v3.0 Integrated With Binding Affinity. Proteins, prot.26825. https://doi.org/10.1002/prot.26825
2. Zhang, O., Naik, S.A., Liu, Z.H., Forman-Kay, J., Head-Gordon, T., 2024. A curated rotamer library for common post-translational modifications of proteins. Bioinformatics 40. https://doi.org/10.1093/bioinformatics/btae444
3. Shapovalov, M.V., Dunbrack, R.L., 2011. A Smoothed Backbone-Dependent Rotamer Library for Proteins Derived from Adaptive Kernel Density Estimates and Regressions. Structure 19, 844–858. https://doi.org/10.1016/j.str.2011.03.019
4. Dunbrack, R.L., Karplus, M., 1993. Backbone-dependent Rotamer Library for Proteins Application to Side-chain Prediction. Journal of Molecular Biology 230, 543–574. https://doi.org/10.1006/jmbi.1993.1170
5. Mukherjee, S., Bahadur, R.P., 2018. An account of solvent accessibility in protein-RNA recognition. Sci Rep 8, 10546. https://doi.org/10.1038/s41598-018-28373-2

---

📄 Citation

If you use these rotamer libraries or codebase in your work, please cite:

Sunandan Mukherjee, Shri Kant, and Ranjit P. Bahadur (2026). Conformational Transitions and Rotamer Preferences of RNA-Binding Proteins Upon Binding RNA.