PROJECT SUMMARY
Recently virus particles have gained immense attention for their potential as transformative therapeutic carriers.
For example, the FDA-approved biologic Luxturna is an adeno-associated virus (AAV) used to carry genetic
materials to treat hereditary blindness. In drug delivery, quality control is very important and the percentage of
empty and part-filled capsids in such a sample needs to be determined accurately to avoid possible catastrophic
effects of overdosing. While there are existing techniques like ELISA and qPCR that are used for this purpose,
blind studies have shown alarmingly high intra-sample standard deviations. We propose a simpler yet more
effective technique to discriminate AAVs depending on their cargo content – a plasmonic nanopore sensor with
automated recapture capability for electrical sensing, aligned to operate in tandem with an optical trapping
system. Viruses are soft nanoparticles and the deformability of AAV capsids depends on their cargo content. We
will implement our bimodal virus characterization platform through our three Specific Aims: (1a) Characterize
deformation dynamics of single AAVs during translocations through solid-state nanopores; (1b) Recapture each
virus after translocation and automate a recapture protocol; (2) Capture single AAVs by self-induced back-action
(SIBA) actuated nanopore electrophoresis (SANE) optical trapping in tandem with electrical recapturing; (3)
Achieve robust classification of AAVs based on their cargo load by applying machine learning to optical-electrical
signals that depend on virus deforamability during experiments. First, we will track the voltage-induced
deformability of three AAV samples of different cargo contents: filled with ssDNA (AAVssDNA), dsDNA (AAVdsDNA),
and empty capsids (AAVempty) through a range of voltages. Each type of AAV is expected to induce a unique and
reproducible change in relative translocation current versus applied voltage. A narrow voltage range that shows
the best discrimination will be applied on different percentage mixtures of AAVssDNA+AAVempty and
AAVdsDNA+AAVempty. Conditions will be optimized to get the best discrimination between cargo-filled AAVs and
the empty/part-filled capsid populations. Once this step is successful, we will incorporate single/multiple
recapture capability to the platform to gain statistically sound data per AAV particle. Next, a gold double nanohole
structure will be fabricated on top of the nanopore to optically trap AAVs by SIBA force, which will further deform
AAVs and slow down their translocation. Importantly, optical-electrical tandem sensing will allow characterization
of single AAV size deformations decoupled from charge effects. For automatic classification of viruses, we will
develop a probabilistic (Bayesian) machine learning model with multiple data segments from optical-electrical
signals of single AAVs for very high accuracy predictions. The proposed machine leanring assisted electrical-
optical sensing will detect the fraction of viral particles (e.g., the AAVdsDNA fraction) from a complex AAV mixture
while using only pico- to nano-molar concentrations, in contrast to current analytical ultracenrifugation
technologies that use up significant amounts of an AAV preparation in the quality assurance step.