ROBUST performs disease module mining (DMM) via the enumeration of prize collecting Steiner Trees (PCST). Given a graph with non-negative edge costs and node prizes, we try to find a subgraph that connets the given nodes to the nodes with the highest prizes, and including the edges with the highest edge weights. The node prizes are initialized using seeds provided by the user, and are updated with the progression of the algorithm.
In mathematical terms: Given a graph \(\mathbf{G=(V,E,c,\boldsymbol{\pi})}\), with non-negative edge costs \(\mathbf{c}\) and node prizes \(\boldsymbol{\pi}\), the PCST-problem tries to find a tree \(\mathbf{T=(V_T, E_T) \boldsymbol{\subseteq} G}\), minimizing \(\mathbf{\sum_{uv \boldsymbol{\in} E_T}c(uv)+\sum_{u \boldsymbol{\in} V\setminus V_T}\boldsymbol{\pi}(u)}\).
In ROBUST-Web, we provide the option of running ROBUST using custom or in-built study bias data. The bias-aware edge weights have been calculated using the following equation:
\( \mathbf{ c(uv)=\boldsymbol{\gamma}\cdot\max\{f(u),f(v)\}+(1-\boldsymbol{\gamma})\cdot\frac{\sum_{u^\boldsymbol{\prime} v^\boldsymbol{\prime} \boldsymbol{\in} E}\max\{f(u^\prime),f(v^\prime)\}}{|E|} } \),
where \(\mathbf{f(u)}\) is the degree of over-representation of protein \(\mathbf{u}\) in PPI networks
and, \(\boldsymbol{\gamma}\) is the extent to which study bias data has been leveraged in the calculation of edge weights.
Alternatively, when no study bias data is selected, the default version of ROBUST with uniform edge costs is implemented, where \(\mathbf{c(uv)=1}\) for all edges \(\mathbf{uv}\).
File type allowed | Data format allowed |
---|---|
|
|
Use case | Error message |
---|---|
No seeds uploaded.
|
‘Seeds cannot be empty’
|
File type allowed | Data format allowed |
---|---|
|
|
Use case | Error message |
---|---|
If ‘Upload custom network’ option selected, but no network uploaded.
|
‘Custom network has to be uploaded!’
|
If columns with zero elements in the first two columns (excluding header), is uploaded.
|
‘Custom network has to be uploaded!’
|
If only one column uploaded.
|
‘Custom network with less than two columns uploaded. Please add two columns.’
|
If 3 or more columns uploaded.
|
No error message: The first two columns retained.
|
File type allowed | Data format allowed |
---|---|
|
At least 2 space-separated columns, and at least one element in the first two columns. The first columns should contain the node name; the second column should contain study bias information pertinent with the node. Note that the second column elements must be of type integer (\(\mathbf{Z}\)).
|
Use case | Error message |
---|---|
If ‘CUSTOM’ option selected in ‘Normalize by’ parameter, but no data uploaded.
|
‘Custom study bias data has to be uploaded!’
|
If columns with zero elements in the first two columns (excluding header), is uploaded.
|
‘Custom study bias data has to be uploaded!’
|
If only one column uploaded.
|
‘Custom study bias data with less than two columns uploaded. Please add two columns.’
|
If 3 or more columns uploaded.
|
No error message: The first two columns retained.
|