Multiscale persistent functions for biomolecular structure characterization

Kelin Xia, Zhiming Li, Lin Mu

Abstract

In this paper, we introduce multiscale persistent functions for biomolecular structure characterization. The essential idea is to combine multiscale rigidity functions with persistent homology analysis to construct a series of multiscale persistent functions, particularly multiscale persistent entropies. Unlike previous topological entropy models, our approach incorporates a special resolution parameter, allowing for analysis across various scales. This method provides a robust framework for conformational entropy evaluation and protein classification, introducing the Protein Structure Index (PSI) to quantify structural regularity.

The Problem: The Arbitrariness of Grid-Based Entropy

Understanding protein folding and interaction requires precise measurements of conformational entropy—a metric of molecular disorder. Traditionally, this is calculated by plotting backbone angles and partitioning the resulting graph into a grid. However, this "grid-based" approach suffers from a fundamental flaw: the entropy value is extremely sensitive to the chosen grid size.

Traditional grid-based entropy is highly sensitive to the choice of grid spacing (b), leading to inconsistent characterization of data clusters (a, c).

As shown in the figure above, if the grid is too fine, every data point appears unique; if it is too coarse, distinct structural features are merged. There is no "natural" scale inherent to the grid method. This research addresses this limitation by using topological data analysis (TDA) to allow the data itself to define its natural clusters and scales.

Methodology: From Geometry to Topology

The proposed pipeline moves beyond discrete geometry into the realm of multiscale topology:

  1. Coarse-Grained Representation: The model focuses on $C_\alpha$ atoms, calculating two primary descriptors: the pseudo-bond angle ($\theta$) and the virtual dihedral angle ($\tau$). These angles capture the essential backbone geometry.
  2. Multiscale Rigidity Functions: Discrete ($\theta, \tau$) points are transformed into a continuous density map using kernel functions. A resolution parameter ($\eta$) acts as a tunable "lens," allowing the structure to be viewed at different levels of detail.
  3. Persistent Homology (PH): A "density filtration" is performed on the map. As the density threshold changes, topological features (clusters) appear ("birth") and disappear ("death"). PH tracks these events in a barcode.
  4. Persistent Entropy and PSI: Using barcode lengths, the authors calculate multiscale persistent entropy. From this, they derive the Protein Structure Index (PSI) based on $\beta_0$ (connected components) to measure structural "regularity."
Coarse-grained protein representation (a, b) and the resulting distribution of pseudo-bond (theta) and dihedral (tau) angles (c).

Key Insights and Results

Topology as a Natural Classifier

By treating clusters of bond angles as topological features, persistent homology identifies "regular" structures like $\alpha$-helices. These features are tightly clustered and thus produce long-lasting (persistent) bars in the barcode, distinguishing them from disordered regions.

The "Mediate Range" Advantage

The study found that while the persistent function method aligns with traditional results at extreme scales (very fine or very coarse), it excels in the "middle" resolutions. In this mediate range, traditional methods often lose structural nuance, whereas persistent functions preserve the underlying topological "shape" of the biomolecular data.

Protein Classification and PSI Validation

The method was validated using a subset of the SCOPe database (~900 proteins). It achieved a clear separation between all-alpha and all-beta proteins using only angle information. Furthermore, the Protein Structure Index (PSI) successfully quantified regularity. In a test of 110 high-resolution proteins, the lowest PSI (highest regularity) was found in a regular $\beta$-sheet (PDB ID: 1GK7), while the highest PSI (lowest regularity) corresponded to a protein dominated by disordered loops (PDB ID: 3ZFP).

Conclusion

Multiscale persistent functions offer a robust, data-driven alternative to traditional grid-based entropy calculations. By incorporating a resolution parameter and leveraging persistent homology, the framework provides a stable way to characterize biomolecular structures and quantify their regularity via the PSI. Future work will explore higher-dimensional topological features ($\beta_1$ and $\beta_2$) and apply these metrics to other complex systems beyond biology.

Summary updated

Abstract

In this paper, we introduce multiscale persistent functions for biomolecular structure characterization. The essential idea is to combine our multiscale rigidity functions with persistent homology analysis, so as to construct a series of multiscale persistent functions, particularly multiscale persistent entropies, for structure characterization. To clarify the fundamental idea of our method, the multiscale persistent entropy model is discussed in great detail. Mathematically, unlike the previous persistent entropy or topological entropy, a special resolution parameter is incorporated into our model. Various scales can be achieved by tuning its value. Physically, our multiscale persistent entropy can be used in conformation entropy evaluation. More specifically, it is found that our method incorporates in it a natural classification scheme. This is achieved through a density filtration of a multiscale rigidity function built from bond and/or dihedral angle distributions. To further validate our model, a systematical comparison with the traditional entropy evaluation model is done. It is found that our model is able to preserve the intrinsic topological features of biomolecular data much better than traditional approaches, particularly for resolutions in the mediate range. Moreover, our method can be successfully used in protein classification. For a test database with around nine hundred proteins, a clear separation between all-alpha and all-beta proteins can be achieved, using only the dihedral and pseudo-bond angle information. Finally, a special protein structure index (PSI) is proposed, for the first time, to describe the "regularity" of protein structures. Essentially, PSI can be used to describe the "regularity" information in any systems.
BibTeX
@article{xia2017multiscale,
  author = {Kelin Xia and Zhiming Li and Lin Mu},
  title = {{Multiscale persistent functions for biomolecular structure characterization}},
  journal = {Bulletin},
  year = {2017},
  doi = {10.1007/s11538-017-0362-6},
}
Figure