Research · Build · DiscoverComputational Science & Engineering

Computational Research

Valley Fever

Structure-guided screening of transporter binding pockets.

Can structural models and short molecular-dynamics simulations help prioritize binding pockets in Coccidioides immitis ABC transporters?

Five predicted ABC transporter structures from Coccidioides immitis, labeled A to E and drawn in blue, with predicted binding pockets highlighted in green for stable, yellow for stable with marginal drift, and orange for unstable
Predicted binding pockets in the five transporters, with examples of stable and unstable binding in 2 ns simulations.

Valley Fever is a fungal infection with about 20,000 reported U.S. cases a year and rising antifungal resistance.

The ABC transporters that may pump drugs out of Coccidioides immitis cells had not been structurally characterized. This study predicts the structures of five of them, docks five known inhibitors into their predicted pockets, and uses short molecular-dynamics simulations to decide which pockets deserve further study.

Key Results

Five C. immitis ABC transporters · five candidate inhibitors

Complexes docked
520
Five transporters, their predicted pockets, and five ligands, docked with AutoDock Vina
Shortlisted for simulation
26
Docking score ≤ −8.0 kcal/mol and pocket probability ≥ 0.20, plus alignment with reference structures
Initially stable pockets
17
Ligand drift under 2.5 Å with contacts retained in 2 ns simulations, run in triplicate

Extended 20 ns simulations of a representative subset confirmed that pockets screened as stable generally persisted. Milbemycin oxime was the most consistent binder.

How It Works

Each stage keeps only the complexes worth studying further.

  1. Select transporters

    Five C. immitis ABC transporters, found by BLASTp search against 23 resistance-mediating reference proteins.

  2. Predict structures

    AlphaFold2 models with a mean confidence score (pLDDT) above 95.

  3. Dock ligands

    PrankWeb pockets and five known inhibitors, docked with AutoDock Vina: 520 complexes.

  4. Static filter

    Docking score and pocket probability thresholds, checked against reference structures: 26 complexes.

  5. Dynamic filter

    2 ns molecular dynamics, keeping low ligand drift and retained contacts: 17 stable pockets.

Scatter plot of docking score against pocket probability for all docked complexes. Dashed lines mark −8 kcal/mol and 0.2; the complexes below and to the right of both lines are colored by transporter, and the rest are gray.
Static filterEach point is a docked complex. Colored points pass both thresholds and were shortlisted.
Line plot of ligand RMSD over 2,000 picoseconds for three simulation replicas. All three rise within the first 250 picoseconds and then stay between about 1 and 2 ångströms.
Dynamic filterLigand drift over the first 2 ns for verapamil in pocket 3 of CIMG_00533. Each line is one replica; the trajectories stabilize within about 2 Å.

Figures from the research poster.

Short simulations support early-stage pocket prioritization; they do not establish long-term stability, functional inhibition, or treatment efficacy.

What I Did

I developed the structure-guided screening workflow, modeled the transporter structures, analyzed the docking and molecular-dynamics results, and wrote the paper as its sole author.

  • Selected five transporters by homology search
  • Predicted and assessed the protein structures
  • Docked 520 protein–ligand–pocket complexes
  • Ran and analyzed the molecular-dynamics simulations
  • Compared pockets with reference transporter structures
  • Wrote the paper and released the code and dataset

Research Paper

Sole-authored paper

Paper record

Code & Dataset

valleyfeverresearch repository

View on GitHub

Research Poster

IEEE BHI 2025 conference poster

View poster

Valley Fever Research Poster

View full-size image (opens in a new tab)