Additional information

DISTRIBUTIONS OVER CHROMOSOMES The two figures below show the distribution of different cellular roles and subcellular localizations for proteins belonging to each chromosome in the human genome. Each spot represents the normalised frequency of the number of proteins (belonging to a certain category) transcribed from a particular chromosome. This number is obtained by dividing the absolute number of proteins (on a spot) by the number of proteins transcribed by the chromosome and by the number of proteins occuring in the category: Normalised freq. = Absolute freq. / (# chromosome * # category)
FEATURES REPRESENTATIONS In ProtFun many different predicted protein features are used as input. These features are represented in a number of different ways: Single value The feature consists of a single value for each protein. Example: The gravy feature consists of one number, the hydrophobicity. Several numbers The feature consists several different scores, the names of which are listed. Example: The signalp feature consists of a fixed number of values. These are the meanS score, the maxY score and the logarithm of the length of the predicted signal peptide (all of which are obtained from SignalP). N bins A prediction method outputs a value per residue. These are encoded as a fixed number of values by dividing each sequence into N bins and averaging over the residue scores within each bin. Example: The phosY feature consists of the average tyrosine phosporylation potential within each of 10 sequence bins of equal size. Several values in N bins Some prediction methods output several values per residue. Each of these values are binned as described above. Example: The psipred feature consists of a score for helix, sheet and coil each averaged within five bins, giving a total of 15 input values. The following table summarize the encoding scheme for all features actually used in the ProtFun method. Abbriviation Encoding Description ec single value Extinction coefficient predicted by ExPASy ProtParam gravy single value Hydrophobicity predicted by ExPASy ProtParam nneg single value Number of negatively charged residues counted by ExPASy ProtParam npos single value Number of positively charged residues counted by ExPASy ProtParam nglyc potential in 5 bins N-glycosylation sites predicted by NetNGlyc oglyc potential-threshold in 10 bins GalNAc O-glycosylations predicted by NetOGlyc pest fraction in 10 bins PEST rich regions identified by PESTfind phosST potential in 10 bins Serine and threonine phosporylations predicted by NetPhos phosY potential in 10 bins Tyrosine phosporylations predicted by NetPhos psipred helix, sheet, coil in 5 bins Predicted secondary structure from PSI-Pred psort 20 probabilities Subcellular location predtions by PSORT seg fraction in 10 bins Low-complexity regions identified by SEG signalp meanS, maxY, log(cleavage pos) Signal peptide predictions made by SignalP tmhmm inside, outside, membrane in 5 bins Transmembrane helix predictions made by TMHMM NEURAL NETWORK ARCHITECTURES In ProtFun an ensemble of five neural networks is used for each predicted category. These are all three layer feed-forward neural nets, that is they have one hidden layer. The number of neurons in this layer is not the same in all networks; the number of hidden units can be found in the table below. The table also shows which features are used as input for each network. Category Hidden units Input features Amino acid biosynthesis 30 ec psipred psort tmhmm 30 ec psipred tmhmm 30 ec netoglyc psipred psort 30 gravy psipred psort 30 oglyc psipred psort Biosynthesis of cofactors 50 gravy pest psort tmhmm 50 ec psipred psort seg 30 ec psipred psort tmhmm 30 gravy psipred psort seg 40 ec psipred psort seg Cell envelope 40 signalp tmhmm 40 nglyc psort signalp tmhmm 30 nglyc psipred psort signalp tmhmm 30 psort signalp tmhmm 30 psipred psort signalp tmhmm Cellular processes 30 gravy nglyc psort seg 30 gravy phosST psipred 30 phosST pest psort seg 30 pest psort seg 30 psort seg Central intermediary metabolism 50 ec nneg npos psipred psort tmhmm 30 nneg npos psipred psort tmhmm 30 nneg npos psipred tmhmm 30 ec psipred psort 30 nneg npos psipred Energy metabolism 10 ec phosST phosY pest psort signalp 40 ec phosST phosY psort signalp 50 ec pest psort signalp 30 phosY pest psort signalp 30 pest psort signalp Fatty acid metabolism 30 gravy phosST pest signalp 30 gravy seg signalp 30 gravy pest seg signalp 30 gravy seg 30 pest psipred seg signalp Purines and pyrimidines 10 gravy nneg npos psort 40 gravy nneg tmhmm 20 gravy nneg npos tmhmm 30 gravy nneg npos psort 30 ec gravy nneg tmhmm Regulatory functions 30 phosST phosY pest psort 30 oglyc nglyc pest psort 30 oglyc phosST phosY pest psort 30 npos nglyc pest psort 40 pest psort Replication and transcription 30 oglyc nglyc psort tmhmm 30 oglyc psort 30 oglyc psort tmhmm 30 nglyc psort tmhmm 30 gravy nglyc psort tmhmm Translation 30 phosY nglyc pest signalp 30 oglyc pest signalp 30 oglyc phosY nglyc signalp 10 oglyc pest signalp tmhmm 10 phosY nglyc pest signalp Transport and binding 40 ec nglyc psort signalp 30 npos psort 40 ec gravy nglyc psort signalp 30 gravy nglyc psort signalp 30 ec psort Enzyme/nonenzyme 40 ec nneg npos psort tmhmm 40 ec npos phosY psipred psort tmhmm 10 ec nneg psort 10 ec nneg npos psort 40 ec psort tmhmm Oxireductase 10 ec oglyc pest psort signalp 30 ec oglyc psort signalp 30 ec gravy phosY pest psort 50 ec phosY pest psort signalp 40 ec oglyc psort signalp Transferase 50 nneg nglyc psort tmhmm 30 nneg phosY psort seg 30 nneg nglyc psort seg 20 nneg phosY psort 50 ec nneg phosY psort tmhmm Hydrolase 50 ec gravy psipred psort signalp tmhmm 30 ec nneg psipred tmhmm 20 ec gravy psipred signalp 30 ec gravy psipred tmhmm 50 ec psort signalp tmhmm Lyase 50 gravy nneg phosST nglyc tmhmm 30 nglyc psort tmhmm 50 nneg npos phosST psort tmhmm 30 gravy nneg phosST nglyc psort tmhmm 30 gravy nneg nglyc psort tmhmm Isomerase 30 nneg npos psipred seg tmhmm 30 nneg pest psipred tmhmm 30 nneg pest psipred seg 30 ec nneg psipred 30 ec npos pest psipred tmhmm Ligase 30 nneg psipred psort 30 pest psipred psort tmhmm 10 nneg psipred signalp tmhmm 30 ec gravy nneg psipred 30 gravy nneg psipred CORRESPONDENCE Lars Juhl Jensen,

This file was last modified Wednesday 29th 2003f October 2003 15:36:56 GMT

Follow CBS on Twitter