next up previous
Next: Clustering using multiple models Up: Definitions Previous: The sequence model,

The sequence-expression model, ${\cal SE}$

Combining equations (2) and (8), we can now write down a joint likelihood for a transcript's promoter sequence alignment and expression profile:


 
$\displaystyle {\mbox{Pr}\left[\bar{S}\vec{E}^{(g)}, a^{(g)} \vert {\cal SE}, \Lambda, \Theta''\right] =}$
    $\displaystyle \mbox{Pr}\left[\bar{S}^{(g)}, a^{(g)} \vert {\cal S}, \vec{p}, {\... ...}\right] \mbox{Pr}\left[\vec{E}^{(g)} \vert {\cal E}, \vec{\mu}, {\bf N}\right]$ (11)

where $\bar{S}\vec{E}^{(g)}$ is the combined sequence-expression information for gene g, $\Lambda = \{ {\bf q}, \vec{\mu} \}$ are the variable parameters, $\Theta'' = \{\Theta,\Theta'\} = \{{\bf N}, \vec{\nu}, {\bf C}, \vec{p}, \vec{D}\}$ are the fixed global parameters and ${\cal SE}$ is our notation for the sequence-expression model.

We can also write down the total joint log-probability for the entire set of transcripts by combining equations (4) and (9):


 
$\displaystyle {{\cal L}\left[{\bf\bar{S}\vec{E}}, \vec{a}, \Lambda \hspace{.3em} \vert \hspace{.3em} {\cal SE}, \Theta'', \vec{\rho}\right] =}$
    $\displaystyle {\cal L}\left[\bar{\bf S}, \vec{a}, {\bf q} \hspace{.3em} \vert \... ...\vec{\mu} \hspace{.3em} \vert \hspace{.3em} {\cal E}, \Theta, \vec{\rho}\right]$ (12)




2000-04-26