BPS2026 – Speeding up a mechanistic model describing the sensory cells of hearing using machine learning

Hearing loss most often arises from damage to the sensory outer hair cells in the inner ear. Outer hair cells employ an organelle, called the hair bundle, to convert sound-induced forces into receptor currents. Outer-hair-cell bundle receptor currents drive the cochlear amplifier, a process that enables us to detect quiet sounds, to hear a wide range of sound intensities, and to have sharp frequency tuning. To better understand the outer-hair-cell bundle, we simulate the mechanical deflection of the bundle and the corresponding receptor current using a mechanistic model. However, simulating a mechanistic model of the bundle for large ranges of parameter values using conventional methods takes a very long time. To speed up the simulations, we train a machine-learning model to capture the behavior of the mechanistic model. The machine-learning model runs orders of magnitude faster than the mechanistic model. Owing to this speedup, we are able to use the machine-learning model to determine which parameters are more important for the function of the outer-hair-cell bundle.