Comparing Structural and Dynamic Differences Among Globin-like Proteins
Yajie HU *, Sandhya P. TIWARI
Institute for Protein Research (IPR), Osaka University
Globin-like proteins share a canonical 3-over-3 α-helical bundle that has remained remarkably conserved throughout evolution, yet the fold supports a spectrum of biological roles that ranges from oxygen transport and photoreception to diverse redox and catalytic chemistries. To clarify how a single structural scaffold can encode such functional breadth, we assembled and curated a comprehensive data set of 819 CATH-classified globin-like structures. After sequence de-redundancy with CD-HIT at 90 % identity and structural filtering with TM-align (TM-score > 0.90), 104 non-redundant proteins comprising 173 chains were retained; oversized multidomain entries were manually trimmed to their globin cores to ensure consistent comparisons.
The study analyzes both geometric and intrinsic‐dynamic signatures. Global geometry is captured through pairwise RMSD, helix–helix angles, lengths, and centroid distances, while intrinsic dynamics are quantified with an elastic-network model to obtain Bhattacharyya-coefficient (BC) similarity matrices and residue-level RMSF profiles. The BC‐based hierarchical clustering partitions the data into ten dynamic groups, reducing noise and revealing clear inter-group differences: proteins whose BC values fall below 0.85, such as 4LMX and 4LM6, exhibit dynamics that deviate strongly from other members despite comparable global folds. RMSF comparisons indicate that these deviations concentrate at specific helices and loop segments and correlate with biological context—group 4 is dominated by dimers whose interface loops are rigidified, whereas group 5 comprises heme-free monomers displaying elevated flexibility near helix F.
Multiple-sequence alignment highlights a conserved structural backbone but fails to expose obvious sequence motifs for the flexible hotspots, implying that functional tuning arises primarily from subtle geometric rearrangements rather than primary-sequence changes. Representative structures such as the paradigmatic 1MBN underscore a complex interplay between conserved scaffolding and variable local dynamics.
Building on these insights, the next phase will integrate RMSF vectors, helix-geometry descriptors, ligand occupancy, and pocket metrics into machine-learning and deep-learning models to predict functional class and to probe how engineered changes in helix length or angle might redirect activity. Planned wet-lab validation—including site-directed mutagenesis and ligand-binding kinetics—will provide a rigorous test of these predictions and illuminate the causal chain linking fold geometry, intrinsic dynamics, and biochemical function.
Through fine-grained structural-dynamics analysis coupled with predictive modeling, this work seeks to explain how a highly conserved fold can diversify into myriad functional niches and, ultimately, to furnish design principles for engineering oxygen affinity, photoreactivity, and catalytic versatility in globin-like proteins.