Bianca: Automated classification of simple (binary) and complex (multiclass) behavior in mice

In the past decade, the field of systems neuroscience has increasingly relied on genetic tools to dissect complex neural circuits. Tools such as optogenetics and pharmacogenetics afford neuroscientists the ability to manipulate specific neural populations to investigate their role in various behavioral processes. A growing problem in the field has arisen from the quantification of such behavioral data. In simple behavioral assays, such as locomotor or feeding behavior, generating an unbiased, repeatable quantification is fairly simple. However, many experimental paradigms attempt to answer more interesting questions about animal behavior and thus output more complex behavioral data. With increased complexity comes increased variation and bias in quantifying this data, leading to poor repeatability across labs. Aiming to alleviate this problem, we developed a program to automate classification of complex behavioral data.