Quantitative Biotherapeutic Particle Characterization via Deep Neural Networks

Project SummaryDespite their many bene ts and common useprotein based drugs can elicit serious adverse side effectsMany of these side effects are believed to be caused by small protein aggregatesOne tool for measuring these
particles in a high throughput fashion is Micro ow ImagingMFIMFI is commonly used in both academia and
industry to characterize subvisible particlesthosem in sizein protein therapeuticsIn many formulations
protein aggregates that arem in size account for upwards ofof the protein aggregates in the productSubvisible protein aggregates have been demonstrated to correlate with adverse drug responseshoweverwhich
speci c protein aggregates induce immunogenic responses remains unknownPharmaceutical companies are required to record and catalog vast volumes of FIM data on protein therapeutic
productsbut are only mandated under FDA regulationsi eUSPto control the number of particles
exceedingandm in delivered productsHence a vast amount of digital images are available to analyzeRecent studies have indicated that some of the factors correlated with adverse drug responses are encoded in
MFI image dataCurrent state of the art MFI analysis methods rely on a relatively low dimensional list ofmorphological featuresto characterize particlesbut these methods ignore an enormous amount of information encoded in the existing large digital image repositoriesDeep Convolutional Neural NetworksCNNs orConvNetshave demonstrated the ability to extract predictive information from raw macroscopic image data without requiring the selection
or speci cation ofmorphological featuresin a variety of tasksHoweverthe heterogeneitypolydispersity of protein therapeuticsand optical phenomena associated with subvisible MFI particle measurements introduce new
challenges regarding the application of CNNs to MFI image analysisThis proposal will spring from state of the art deep CNN methods to provide new analysis tools capable of
reliably analyzing and classifying heterogeneous MFI protein therapeutics dataThe envisioned software productcapable of processing images from both of the leading manufacturers of MFI instrumentsFluid Imaging Incand
ProteinSimplewill provide high throughputdata driven models that ef ciently capture information encoded in
the large collection of image dataavoiding the need to de nefeaturesa priori and is anticipated to provide
a paradigm shift to the MFI quanti cationeldWe anticipate that the proposed algorithms and software will
help in correlating which protein aggregates induce adverse side effects and will also serve as a useful process
monitoring tool Project NarrativeProtein aggregatesfound to some level in all commercial protein therapeutic based drugscan be associated with a number of adverse responses in patientsIn this projectnew machine learning algorithms will
be developed for analyzing large collections of biotherapeuticsprotein therapeutic based drugsmeasured via
micro ow imagingMFImicroscopyPhase I will develop prototype software capable analyzing images of protein
aggregrates in biotherapeutics measured via MFI for both process monitoring and identifying which characteristics of protein aggregates are correlated to high risk for causing adverse responses in patients