Modeling Feasible Locomotion of Nanobots for Cancer Detection and Treatment

Noble C. Harasha, Cristina Gava, Nancy A. Lynch, Claudia Contini, Frederik Mallmann-TrennProceedings of the National Academy of Sciences 2025

Introduction

Targeted drug delivery is often described as the "holy grail" of oncology. Traditional chemotherapy is systemic, circulating through the entire body and damaging healthy tissue alongside cancerous cells. Nanobots—engineered particles at the billionth-of-a-meter scale—offer a potential solution by navigating directly to a tumor to release their therapeutic payload. However, at the nanoscale, physics presents unique challenges: agents lack traditional power sources and are subject to constant, noisy buffeting known as Brownian motion. This paper proposes a mathematical framework to bridge the gap between theoretical swarm robotics and the physical reality of nanoparticles, providing a formal model for how these machines can autonomously find a target in a colloidal environment.

The Orientation-Biased Movement Model

The researchers developed the Orientation-Biased Movement Model, which treats nanobot locomotion as a "biased random walk." Instead of a perfectly steered path, the direction of each step is drawn from a truncated normal distribution. The "bias"—the likelihood of the bot pointing toward the tumor—is determined by the steepness of a chemical gradient, mimicking biological chemotaxis.

The study explores two primary model variants:

  1. Passive Agent Model: This assumes the tumor naturally emits a steady chemical signal (an endogenous gradient) that the bots follow.
  2. Active Agent Model: This utilizes a "swarm intelligence" approach. In this scenario, a portion of the bots carry a signaling chemical. When these "first responders" locate the tumor, they release the chemical to create and amplify a beacon, significantly strengthening the gradient for the rest of the swarm.

Methodology and Validation

As a theoretical modeling study, the framework was validated using a combination of stochastic analysis and empirical data. The authors used majorization and stochastic dominance to derive analytical upper bounds on "hitting time"—the duration required for 75% of the swarm to reach the target. To ensure the model's locomotion parameters were physically grounded, they were tuned against experimental data from Joseph et al. (2017), which tracked the movement of chemotactic synthetic vesicles in vitro. Simulations were conducted in 2D Euclidean spaces representing colloidal environments like blood or interstitial fluid.

Dynamics of Swarm Navigation

The research identified that a nanobot’s journey toward a tumor consists of three distinct phases based on its distance from the target:

  • The Distal Phase: Far from the target, the chemical signal is too weak to provide direction, and the bot moves randomly.
  • The Intermediate Phase: In this "sweet spot," the gradient is at its steepest, and the bot's movement becomes highly directed toward the source.
  • The Proximal Phase: Very close to the tumor center, the gradient flattens out again, and the bot reverts to a noisy search to find an exact binding site.

A critical finding was the "First Responder" Effect in the Active Model. The total runtime is largely determined by the time it takes for the first bot to find the tumor by chance. Once that bot releases its signal payload, the hitting time for the remainder of the swarm drops by three orders of magnitude. This mechanism also helps combat natural diffusion, as successive agents arriving at the site "refresh" the chemical beacon.

Key Results and Performance

The proposed models significantly outperformed simple random walks. The Active Model, in particular, demonstrated a 1,000x improvement in runtime performance as swarm size increased. The authors mathematically proved that for a swarm of size $n$, at least 75% of agents will complete their delivery by a specific time $10t^+$ with a probability of at least $1 - 2^{-n/4}$. However, the study also noted that hitting times increase non-linearly with the initial distance from the tumor, suggesting that nanobots should be injected as close to the suspected site as possible for maximum efficacy.

Limitations and Future Directions

While the model provides a robust theoretical foundation, several real-world factors require further study:

  • Dimensionality: The current model is 2D; future work must account for the 3D complexity of the human body.
  • Fluid Dynamics: The model assumes a relatively still environment, but actual blood flow (advection) would introduce significant noise and directional forces.
  • Biological Clearance: The body’s immune and filtration systems (renal and hepatobiliary) constantly clear foreign particles. The "Active" signal amplification must occur faster than the rate of clearance to be effective.
  • Metastasis: Future iterations will need to address how a swarm can effectively divide itself to treat multiple metastatic sites simultaneously.

Summary updated

Abstract

Deploying motile nanosized particles, also known as "nanobots," in the human body promises to improve selectivity in drug delivery and reduce side effects. We consider a swarm of nanobots locating a single cancerous region and treating it by releasing an onboard payload of drugs at the site. At nanoscale, the computation, communication, sensing, and locomotion capabilities of individual agents are extremely limited, noisy, and/or nonexistent. We present a general model to formally describe the individual and collective behavior of agents in a colloidal environment, such as the bloodstream, for cancer detection and treatment by nanobots. This includes a feasible and precise model of agent locomotion, inspired by actual nanoparticles that, in the presence of an external chemical gradient, move toward areas of higher concentration by means of self-propulsion. We present two variants of our model: the first assumes an endogenous chemical gradient fixed over time and centered at the cancer site; the second is a more speculative, dynamic variant in which agents themselves create and amplify a gradient centered at the cancer site. In both settings, agents sense the gradient and ascend it noisily, locating the cancer site more quickly than by simple Brownian motion. For the first variant, we present simulation and analytical results to bound the time it takes for agents to reach the cancer site. For the second, simulations highlight collective benefit from agent-generated signaling, showing significant runtime improvement via chemical signal amplification.
BibTeX
@misc{harasha2025modeling,
  author = {Noble C. Harasha and Cristina Gava and Nancy A. Lynch and Claudia Contini and Frederik Mallmann-Trenn},
  title = {{Modeling Feasible Locomotion of Nanobots for Cancer Detection and Treatment}},
  year = {2025},
  eprint = {2507.12400},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2507.12400},
}
Figure