neural dynamics
causal protein networks
my undergraduate thesis with the ucla neural dynamics group. it defines a functional interactomes score (fis) for protein pairs, computes it from published experiments, and assembles the results into a directed network.
databases like string estimate how likely two proteins are to interact. fis asks a different question: did one protein measurably change the other, and in what direction. that answer becomes the edge weight.
sections
1 of 5
why causal edges
biophysical models do not scale, and predictive networks still require validation.
models like hodgkin-huxley describe a neuron precisely but cannot absorb arbitrary new molecular components. network models scale well, but their edges are confidence estimates built from co-occurrence and co-expression, so each one still needs experimental confirmation.
fis takes the middle path. nodes are proteins and each directed edge represents a measured functional change that one protein induced in another, so the graph is built from evidence rather than likelihood.
arrow keys move between sections.
the full thesis covers the methods, the fis derivation, and every figure.
ucla computational and systems biology, 2023. mentor: dr. sharmila venugopal.