Weighted persistent homology for biomolecular data analysis

Zhenyu Meng, D. Vijay Anand, Yunpeng Lu, Jie Wu, Kelin Xia

TL;DR

Researchers have developed Localized Weighted Persistent Homology (LWPH), a new topological framework that decomposes complex biomolecules into local, overlapping domains to capture fine-grained structural nuances. By assigning physical and chemical weights to atoms and bonds, this method successfully discriminates between DNA types and identifies subtle configurational states in ionic liquids that traditional global models miss.

What Problem Does This Paper Address?

Biomolecular data is exploding in volume and complexity, with millions of sequences in GenBank and over 150,000 structures in the Protein Data Bank. To make sense of this, researchers often turn to Topological Data Analysis (TDA), specifically Persistent Homology (PH). PH is excellent at identifying the "shape" of data—like holes, loops, and voids—across different scales.

However, traditional PH has two major "blind spots" in biology:

  1. Homogeneity: It often treats all atoms as identical points in space, ignoring crucial chemical properties like atom type, partial charge, or bond strength.
  2. Global Bias: It typically analyzes a molecule as one inseparable system. In large molecules like DNA or proteins, global topological features can drown out the subtle, local structural changes that actually drive biological function.

This paper seeks to bridge the gap between abstract topology and concrete molecular physics by making TDA both "weighted" (physics-aware) and "localized" (spatially-aware).

Methods

The authors first provide a systematic review of Weighted Persistent Homology (WPH), where weights are assigned to vertices (atoms), edges (bonds), or higher-order structures. These weights can represent physical properties like atomic radii or chemical properties like interaction potentials.

The core innovation is Localized Weighted Persistent Homology (LWPH). Instead of calculating the topology of the entire molecule at once, the authors:

  1. Decompose the structure: They break the molecule into a series of local domains (represented as overlapping spheres).
  2. Apply WPH locally: Within each domain, they apply a weighted distance matrix. For DNA, they specifically use a weight definition that sets the distance between atoms in the same residue to zero, effectively forcing the model to focus on non-covalent interactions between different residues.
  3. Generate Fingerprints: The results are visualized as "Betti curves" or Persistent Betti Numbers (PBNs), which serve as topological fingerprints for local regions.

Datasets

The study evaluates the model using two primary sources:

  • Static DNA Structures: 30 PDB files representing three major DNA configurations: A-DNA, B-DNA, and Z-DNA (10 files per type).
  • Molecular Dynamics (MD) Trajectories: Simulations of a 16-mer DNA duplex in two different environments:
  • Water (WAT): A standard aqueous environment.
  • Ionic Liquid (IL): A more restrictive environment consisting of $[BMIM]^+$ and $[BF_4]^-$.
  • The MD data consisted of 5,000 sample frames extracted from the last 10ns of 6 different simulation trajectories.

Key Insights and Ideas

  • The Power of Localization: The authors found that global PCA models (based on atom coordinates or standard PH) completely failed to cluster DNA configurational states in ionic liquids. Only when they switched to the localized model (LWPH) did the distinct structural states emerge clearly.
  • Topology vs. Geometry: Traditional "helical coordinate" systems are the gold standard for DNA analysis, but they are mathematically complex and usually limited to one or two base pairs. LWPH provides comparable accuracy to these geometric models but is much more flexible, allowing researchers to analyze domains of arbitrary size and shape.
  • Capturing "Confinement": The model quantitatively demonstrated the "confinement effect" of ionic liquids. DNA in ionic liquids showed a much smaller distribution area in PCA space compared to DNA in water, confirming that the ionic environment significantly restricts the molecule's structural fluctuations.

Key Results

  • DNA Classification: LWPH-based features successfully discriminated between A-, B-, and Z-type DNA. While standard Local Persistent Homology (LPH) struggled to find clear boundaries, the weighted version (LWPH) showed distinct clusters in PCA plots.
  • State Identification: In ionic liquid environments, LWPH identified two distinct configurational states of DNA. These states correspond to different ion-DNA binding modes that are invisible to global topological models.
  • Quantitative Consistency: The "area of distribution" (a measure of conformational fluctuation) was significantly lower in ionic liquids. For example, using LWPH-PCA, the distribution area for DNA in ionic liquid (IL2) was 296 units, compared to 554 units in water (Wat1), representing a nearly 50% reduction in fluctuation.

Limitations and Future Work

While the paper demonstrates LWPH's superiority over global models, it notes that the choice of "local domains" and "weight functions" requires domain expertise. The current study focused heavily on DNA; the authors suggest that future work should extend this to proteins, biomolecular complexes, and larger assemblies like chromatin.

Additionally, because LWPH features are "unit-free" and capture local variations so well, they are prime candidates for integration into machine learning and deep learning pipelines for drug discovery and protein folding predictions.

Summary updated

Abstract

In this paper, we systematically review weighted persistent homology (WPH) models and their applications in biomolecular data analysis. Essentially, the weight value, which reflects physical, chemical and biological properties, can be assigned to vertices (atom centers), edges (bonds), or higher order simplexes (cluster of atoms), depending on the biomolecular structure, function, and dynamics properties. Further, we propose the first localized weighted persistent homology (LWPH). Inspired by the great success of element specific persistent homology (ESPH), we do not treat biomolecules as an inseparable system like all previous weighted models, instead we decompose them into a series of local domains, which may be overlapped with each other. The general persistent homology or weighted persistent homology analysis is then applied on each of these local domains. In this way, functional properties, that are embedded in local structures, can be revealed. Our model has been applied to systematically study DNA structures. It has been found that our LWPH based features can be used to successfully discriminate the A-, B-, and Z-types of DNA. More importantly, our LWPH based principal component analysis (PCA) model can identify two configurational states of DNA structures in ion liquid environment, which can be revealed only by the complicated helical coordinate system. The great consistence with the helical-coordinate model demonstrates that our model captures local structure variations so well that it is comparable with geometric models. Moreover, geometric measurements are usually defined in local regions. For instance, the helical-coordinate system is limited to one or two basepairs. However, our LWPH can quantitatively characterize structure information in regions or domains with arbitrary sizes and shapes, where traditional geometrical measurements fail.
BibTeX
@inproceedings{meng2020weighted,
  author = {Zhenyu Meng and D. Vijay Anand and Yunpeng Lu and Jie Wu and Kelin Xia},
  title = {{Weighted persistent homology for biomolecular data analysis}},
  booktitle = {Scientific},
  year = {2020},
  doi = {10.1038/s41598-019-55660-3},
}

Loading related content...

Figure