Das offers two examples. First, machine learning can characterize how water molecules interact through hydrogen bonding, including how water–water hydrogen bonding differs inside a polymer brush layer compared with outside it. Second, when an electroosmotic flow does not behave as researchers expect, machine learning can quickly identify the cause. Without it, Das notes, combing through a trillion data points one by one would take an enormous amount of time.
Das also sees machine learning as a design tool. Combined with all-atom MD, it could help researchers reverse engineer channels with specific behaviors. For example, a channel lined with polymer brushes might be engineered to slow or stop flow. That would allow it to function as a sensor that detects certain particles or moves different particles at different velocities.
This kind of engineering is possible with all-atom MD alone, Das acknowledges, but the timelines differ. Work that would take two weeks with machine learning could take a year without it, because researchers would need to sift through the full data set. Machine learning tells researchers they can skip that step and offers guidance on how to analyze the data. That shortcut, Das says, is among the technique’s biggest contributions.
Further Reading
(1) Das, S.; Workman, Jr., J.; Wetzel, W. Siddhartha Das Discusses Electroosmosis in Soft Nanochannels at SciX 2026. Spectroscopy Online, 2026. https://www.spectroscopyonline.com/view/siddhartha-das-discusses-electroosmosis-in-soft-nanochannels-at-scix-2026 (accessed October 13, 2026).