Biological Computations

Genome Annotation Paper

Mon, 21 May 2007 13:34:42 GMT

Other news my genome annotation paper is progressing well, still in talks with the JGI as to when and how we publish it.

PMF Consensus Profiles For Cross Species Proteomics

Mon, 21 May 2007 13:30:37 GMT

Well after collecting MALDI-TOF spectra last week from my "Zoo Gels" (1d SDS-PAGE Gel electrophoresis of "Rat", "Chinchilla", "Vole", "Pig", "Sheep", and "Goat" muscle protein samples) I have begun compiling the results into consensus spectra.

Looks like I can only really do it for "Creatine Kinase" at this stage because I didn't get many Mascot protein identifications, mainly because there are not many proteins sequenced for most of these species and that the samples where quite old and I'm not very sure that the spectra are of any usable quality.

Anyway it does seem to work which could mean that I will be over in Liverpool collecting lots of Peptide Mass Fingerprint data in the near future.

Joint BSPR / EBI Proteomics Conference 2007

Mon, 21 May 2007 13:22:33 GMT

I have decided to attend this years
Joint BSPR / EBI Proteomics Conference 2007 in Hinxton Cambridge (not a massivly exciting location but lots of interesting speakers)

Ruedi Aebersold is speaking which should be very interesting as I referenced a lot of his work in the review paper I wrote for CCHTS.

Rob will fund me but I could do with trying to find some travel grants from somewhere...

Update

Thu, 17 May 2007 10:01:02 GMT

Right I really need to start using this account more and try and keep upto date with my work. Will come in very useful for thesis writing.

So in the world of Proteome Informatics, I have a review paper comming out sometime soon in the Journal of Combinatorial Chemistry and High Throughput Screening (CCHTS) not a massivly prevalent journal but its my first publication so its just the start. The review its self is quite interesting providing a good introduction to the field of proteome informatics as well as discussing some of the very lastest developments to have come about (namely proteotypic peptide prediction, miss cleavage prediction and new peptide scoring systems).

In other news am writing a research paper about using proteomics data to annotate genomes, and have a new project building consensus spectral libraries.

Anyway goto run am booked onto the MALDI-TOF now to get some more PMF data.

GENOMES to SYSTEMS Manchester 2006

Mon, 27 Mar 2006 11:51:32 GMT

Well I spent most of last week at the Genomes to Systems conference in the GMEX here in Manchester. There was a good turnout about 900 people quite a few I knew as well - everyone from the exUMIST/Manchester Bioinformatics, the guys from Protein Functions Group in Liverpool, a few people from my Post Genomes MSc course, the guys from GAPSIA and even a few of my colleges from AstraZeneca.

In general there was a lot of interesting talks although as always happens at these events alot of the talks went a bit off topic but there was a lot of choice in what I went to see.

Day 1 - Wednesday 22nd March

Opening Lecture by Bernhard Palsson (San Diego University) - "The Human Metabolic Network"
An interesting talk on how they had been rebuilding the the human metabolic network using two different methods - a bottom up method which basically uses high througput methods (search a whole genome analysis methods) to identify lots of metabolites, their associated proteins, sub cellular location and their interactions, generally very inaccurate but is good for discovery of new metabolic pathways. Secondly they used a top down approach in which they compile published and validated data to create a network that is confirmed experimentally and has no errors. Was filled with lots of complex diagrams of pathways.

The was then the choice of lecture sessions either: Structural Genomics of Disease (Chaired by Samar Hasnain) or Proteomics for Systems Biology (Chaired by Rob Beynon [my primary supervisor!!])
So I went to the Proteomics session:

A. Brian Chait (Rockefeller, New York) - "Proteomic tools for Dissecting Cellular Processes"
This lecture ended up being more structural than proteomics but was a quite interesting look a the transport proteins that cross the nuclear membrane in cells (Then central bit that contains the DNA)

B. Alistair Hetherington (Lancaster Uni) - "The Use of Proteomics to Dissect the Plant Guard Cell Signalling Network"
Again not much in the way of proteomics and twas plant biology..! Was not that bad in the end though and I do have some interest in cellular signalling as I did my masters project on protein phosphatases which are used to degrade cellular signals

C. Al Burlingame (San Francisco Uni) - "Toward Defining Epigenetic Protein Signatures in Cellular Networks by Electron Capture Dissociation (ECD) and Methods of Tandem Mass Spectrometry"
By far the longest title of any talk!! A good example of the application of ECD and MS/MS in systems biology

D. Simon Gaskel (CISB, Manchester Uni was head of my Post Genome Masters Course) - "Convergent Analytical Stratagies for Quantitative Proteomics"
Simon gave a good summary of the various technologies that are being applied in Mass Spectrometry in Manchester, protein labelling and post translational modifications as well as co-analysing proteins to compare relative abundance in a sample

LUNCH - not very nice so James K and I went to Pizza Hut for all you can eat!!

Another choice of sessions here: Metabolomics in Health and Disease, Advances in Biosensors and instrumentation, Comparative and enviromental genomics in action
I didn't really have much I wanted to see in these sessions so I wandered from session to session, the biosensors lecture seemed to be mostly sales lectures by companies trying to promote their new instruments, there was a really good talk on "Metabolic profiling in Diabetes" Chris Newgad (Duke) which showed how as rats put on weight and became obese their beta-islets (the clusters of cells that produce insulin in the pancreas) replicated and grew in size to try and compensate for increased sugar intake but after doing this for so long they seemed to loose control and fall apart hence to on set of diabetes as the cells couldn't produce enough insulin any more. Several other lectures went into the idea of personalised medice based on a metabolite analysis - quite and interesting future for medical treatments

POSTER SESSION and SHOWCASES

Closing Lecture - Julie Ahringer (Cambridge University) - "Global Analyses of Gene Function using RNAi"
A very good lecture on the use of RNAi (Short strands of RNA that are designed to complement the mRNA for a particular gene and interfere with the production of a particular protein via the translation process) to systematically interfere with each protein in c.elegans (Worms) and classify each protein as lethal or by the phenotypic (appearance) effects and hence try to work out which proteins work together and what their function is.

EVENING MEAL AND NETWORKING EVENT

Day 2 - Thursday 23nd March

Opening Lecture by Michael Snyder (Yale University) - "Genomes and Proteomes"
An interesting summary of how we can intergrate all the various data in to large scale networks linking proteins with genes and their regulation at various levels

First Parralel sessions: Disease Proteomics, High-throughput Functional Genomics and Gene Manipulation, Public Population projects in Medicine
I attended the Disease Proteomics Session chaired by Simon Gaskell

A. Ron Heeren (Amsterdam Uni) - "Mass Spectrometric Imaging for Proteome Analysis"
This was one of the best talks I attended, disscussing how MALDI (Matrix Assisted Laser Desorbtion Ionisation - The instrument I have been trained to use over in Liverpool) mass spectrometers can be used as microscopes to image cells and tissues on a molecular level. Gave some examples of its use of brain tissue!!

B. Albert Heck (Utrecht) - "A Quantitative Proteomics approach to Drosophila Early Embryonic Development and Embryonic stem cell differentiation"
A great talk on manipulation of stem cells, not much on proteomic methods but had some cool pictures of Drosophila (Fruit Flies) that had various protein stimulated in various stem cells during the embryonic stage and had grown multiple eyes all over their bodies..even on their knees!!!

C. Pierre Thibault (Montreal Uni) - "Comprehensive Detection of Unlabelled Peptides for Targeted Identification of Key Cell Signalling Events"
A lecture on various methods for identifying relative abundance on peptides without use of a radioactive or mass label again focussed of kinases and phosphatases

D. Tony Whetton (Patterson Cancer Research, Manchester Uni) - "Normal and Malignant Stem Cell Proteomics"
Tony discussed the major diffences in the proteomes of normal and malignant stem cells and how cancers could be treated by targeting malignant stem cells

LUNCH - better today (sweet and sour veg, rice + salad)

EMBL-EBI - Uniprot demonstration - latest improvements to Swissprot and TrEMBL protein sequence databases

2nd parallel session - Engineering Approaches to Biological systems, Pharmacogenomics - personalised medicine, Progress in Standards for Reporting 'Omics Data'
I attended the Progress in Standards for Reporting 'Omics Data'

A. Chris Taylor (EMBL-EBI, European Bioinformatics Institute) - "A coordinated Multi-'omics view of data standards"
A disscusion of how the EBI are setting up and implementing data standards for proteomics and Metabolomics in the hope of making data from different labs and instuments more comparable

B. Jason Snape (NERC, Astrazeneca SHE) - "Importance of data standards to facillitate the uptake of Genomic science outputs into science based policy and regulation"
Mainly focussed on Astrazeneca buisness implementation of research and how they regulated their data

C. Charlotte Capener (BBSRC) "The BBSRC data-sharing policy"
A lot about how the BBSRC decide on funding allocation

D. Herbert Thiele (Bruker Daltronics, Bremen) - "The HUPO Brain Proteome Project"
A interesting and relevant talk about automated proteomics and reverse database searching using various software

POSTER SESSION

Closing Lecture - Matthias Uhlen (Stockholm) A Human Protein Atlas for Normal and Cancer cells - Biomarker discovery by Antibody-based proteomics

Day 3 - Friday 24th March

No opening lecture!

parallel session- From Models to cells, Nanoscale Technologies in bioscience, Transcriptomics and Expression Microarrays A promise Delivered?
Again didn't want to see anything in particular so just wandered from lecture to lecture some interesting stuff in the nanoscale technologies lectures.

Closing Lecture Denis Noble (Oxford) "Principles of Systems Biology Using the virtual Heart"
A totally awesome lecture describe how they have created a computational model of a virtual heart on the cellular level modelling all the electronic signals and transport proteins and well as the muscle contraction twas very impressive

LUNCH - Packed lunch

END

All in All twas a great seminar look forward to its next incarnation in a few years time.

Relative protein sequence shannon entropy

Tue, 21 Mar 2006 13:48:08 GMT

So my 6 sets of protein profiles are almost built after fighting with the cluster to get it to run everything. (I have clustered proteins from two different protein sequence datasets, one containing 5 yeast species and one containing 20 vertebrate species including human , I have used three different methods to cluster each of the datasets, PSI BLAST, MCLTRIBE, and PFAM . Hence 6 sets of profiles. Each cluster is then aligned using MUSCLE)

I am now looking at how to compare different protein clusters from a peptide (digested protein fragment) point of view, I have decided to calculate 3 statistics for each profile.
For each cluster I will select a QUERY protein which will act as anchor for all my calculations
(I will probably use the Aspergillus Fumigatus proteins in the yeast data and the Human proteins in the Chordata data)

1. Split each protein in a cluster on lysine (K) and arginine (R) amino acids (as if being digested by the proteolytic enzyme trypsin) .
This will divide each protein into a number of peptides I will calculate the standard deviation of the peptide number from the Query Protein.

2. I will then calculate the average RELATIVE SEQUENCE ENTROPY for each peptide. This is a calculation based on information theory which can be used to score how well each position in a sequence is conserved between all the proteins in a cluster. I have written a script to build a matrix of amino acids frequencies by looking at their occurrence in all the protein sequences in the swissprot database ( A massive amount of sequences) this matrix is then used to adjust normal Shannon entropy (which is the sum of the log(base2) position frequencies) by dividing each log by the natural frequency of the amino acid

H = - SUM Pi Log2 Pi / Qi

where Pi = the fraction of amino acid i at current position and Qi is the natural frequency of amino acid i
which is SUMed for each of the 20 amino acids

if H >= 2 this shows that the position is VARIABLE and will be marked as a V
if H <= 2 this shows that the position is CONSERVED and will be marked as a C
if H <= 1 this shows that the position is HIGHLY CONSERVED and will be marked as a H
(litwin et al, 1992)

3. I will also calculate the relative sequence entropy for each point using a fixed window size, rather than the peptides, which will parse through the protein sequence incrementing at a set step size.

Have not yet thought of the best way to then compare all this data as I will have about 3000+ clusters for each of the 6 profile sets... I think some scatter charts and histograms may be they way fowards at that point.

Binary Genertics

Tue, 07 Feb 2006 12:04:48 GMT

I was sitting getting frustrated at the damned beowolf cluster (very cool specs: 32 node beowolf cluster of dual 2gig opterons with lots of memory and storage) that doesn't work (probably because very cheesily it is called 'agent.smith'),

anyway I was considering how the way that we currently play with nucleotides and genes is very similar to what it must have been like trying to program the very first computers using binary, very slow and low level. Which makes me think that the future is going to be genetic programming!! computer science already models many of its ideas for neural networks and evolutionary computing on biological systems, so what if we did the reverse and applied the concepts of programming to genetics.

I a way our DNA is like a computer program, each gene being an object (each gene having its own properties and features, along with its own list of operations denoted by the promoter and regulatory sequences that different types of enzyme bind to). Which means in theory you could by understanding the way in which genes work and how this effects the proteins (and the function of said proteins) which are encoded by genes write a higher level programming language based on codons and promoters and sequence words (Short sequences of nucleotides that serve a purpose eq. restriction (cutting) site) which could be considered the equivalent of assembly code, and then create and even higher level language which could be used to design and program simple evolving life forms on a computer with specific adaptions for a particular task, almost like they already do with the programmed bacteria PROGRAMMED BACTERIA add this to the recent steps foward in training brain cell for specific tasks (Rat Brains flying F11 jets) and we pretty much have everything you need for a totally organic based technology that is massivly versatile from organic transport (living spaceships) to an organic house. Tis so very crazy, fills me with renewed energy to battle once again with agent.smith and finally submit my alignment MATRIX.....!

Insilico Mass Spectrometry

Wed, 25 Jan 2006 11:29:36 GMT

Have recently had the IT people install a great perl modual for proteomics for me

InsilicoSpectro

Have made use of the digest function so far and its proving to be very useful.
Although it did take a lot of messing about with ENV variables to get it fully working and I am getting and error for one of my sequences which I haven't quite figured out yet.

Genomes to Systems Conference

Tue, 24 Jan 2006 14:45:21 GMT

Improtant conference of the year taking place in Manchester presented by "The Consortium for Post-Genome Science"

GENOMES TO SYSTEMS CONFERENCE

Am already registered, don't think I will submit an abstract though.

Current Work

Tue, 24 Jan 2006 14:30:49 GMT

Well lets start with a quick summary of my current PhD work which is split between Liverpool University and Manchester University.

I am researching computational solutions for cross species protein identification in proteomics.

Although many genomes are now completed, they define a vanishingly small part of the entire variety of life on the planet. Future large scale genome sequencing projects are intended to deliver complete genomes for plants and animals of economic importance, or for those that map, albeit thinly, planetary biodiversity. It follows that a barrier to effective use of proteomics in non-model species is the lack of availability of actual or inferred protein sequence data. The scientific community cannot, and should not, only work on model organisms, and it behoves proteomics experts to discover the most effective ways to use existing proteomics data sets to discover the identities and functions of proteins from other species. We believe that an important contribution to proteomics research in environmental sciences will be the development of a series of experimental and informatics enhancements that greatly enhance the success of cross-species matching.
Peptide mass fingerprinting (PMF) is a straightforward, sensitive and efficient strategy for protein identification. However, it is often claimed that PMF is limited in matching peptide mass spectra from one species against database entries in other species [1-3]. We (McDonald, Beynon et al, in preparation) have just completed a comprehensive analysis of the success of cross-species matching in peptide mass fingerprinting, basing our analysis on a series of skeletal muscle proteins from a range of over 30 animal species. Our goal was to discover the limitation of the methodologies, driven by our need to conduct proteomics projects in wild species without systematically sequenced genomes. From this study, the following general observations emerged. Although, as expected, not every putative peptide was present in the mass spectrum, for each species, the same peptide, albeit of slightly different mass, was present in the mass spectrum. Low levels of amino acid substitutions do not therefore influence the ability of the peptide ion to be detected in the MALDI-ToF instrument. Secondly, when comparing protein sequences for the same protein for different species, some peptides were much more highly conserved than others.
We have already implemented profile-based PMF identification techniques (Lester, Cole, Hubbard, ISMB Conference proceedings 2004) which demonstrate two important points: firstly, that profiles provide a gain in the number of overall matched peptides in a cross-species context compared to random (incorrect) profiles, and secondly, that simple profile-based scoring outperforms single orthologue searches. However, more sophisticated scoring algorithms are required to enhance protein identification. We therefore propose a weighted PMF approach that is tuned for, and tolerant to, cross species matching. First, we propose to modify the existing algorithm by building peptide-specific weighting for the peptides in a profile. Secondly, we will consider different weights for the different parent proteins. This “phylogenetically tuneable” weighting, may be user-modified (since the user knows the species with which they are working) to reconstruct the weight matrix according to phylogenetic proximity. Finally, the match should also be weighted to reflect the propensity of the peptide to be present in the MS – propensity weighting. This eliminates very small or very large peptides, but at the same time, scores more highly those peptides that have a strong representation in the true mass spectrum.
Whilst the phylogenetic weighting can be built from families of aligned sequences, the propensity weighting can be built in one of two ways. First, true experimental data has a significant role to play. However, at present, there is a tendency not to publish these data, and they are rarely captured. However, the Manchester/EBI-based PEDRo groups are pushing data standards forwards and we anticipate that much more PMF and MS/MS data will be accessible as International data repositories come online. Data from the commonly used search engines (MASCOT, ProFound, PepMapper – all already available to the consortium) will be harvested to retain a high quality learning set. In time, this learning set will reflect the experience of large numbers of laboratories. This dataset will in turn allow us to devise predictive algorithms that are capable of assessing the propensity of peptides to be present in a typical MALDI ToF mass spectrum – we may then be in a position to move to a predictive model. We will also continue to generate data sets under more controlled conditions, for this, we will need to ascertain the true origin of individual peptides, validated by tandem MS using a MALDI source. We anticipate that this approach will be generic and will also enhance, same–species PMF identifications. We will develop a new PMF algorithm that includes peptide-specific, phylogenetic and propensity-based peptide weighting and which yields as a putative hit a protein family ID.
[1] Verrills et al., (2000) Cross-matching marsupial proteins with eutherian mammal databases: proteome analysis of cells from UV-induced skin tumours of an opossum (Monodelphis domestica). Electrophoresis. 21:3810-3822.
[2] Lester PJ, Hubbard SJ. (2002)Comparative bioinformatic analysis of complete proteomes and protein parameters for cross-species identification in proteomics. Proteomics. 2:1392-1405.
[3] Liska et al., (2003) Expanding the organismal scope of proteomics: cross-species protein identification by mass spectrometry and its implications. Proteomics. 3:19-28.
Training
The student will share their time between two research groups: those of Prof Beynon in Liverpool and Dr Hubbard in Manchester. The training that they will receive will be cross-disciplinary, and will entail:

Capabilities in sample preparation and PMF data acquisition, across a broad range of species. It is timely and appropriate to extend the data set across a broader phylogenetic base. Training in protein separation/proteome simplification, MALDI-ToF MS for PMF, and tandem MS for ambiguous identification.
Analysis of the relationship between intensity of an ion on MALDI-ToF MS and peptide composition and sequence. The application of bioinformatics, specifically machine learning methods (including GAs and neural networks) to build an improved predictive algorithm.
Bioinformatic analysis of existing data sets to provide new weighting algorithms for degree of conservation and propensity to deliver strong ion currents in MS.
Elaboration of existing and publicly accessible search tools (Pepmapper) to include weighting factors sympathetic to cross-species matching.

I am currently looking at the use of a protein/peptide profile database to improve cross species peptide mass fingerprinting. I have begun testing different clustering algorithms to build these profiles:

PSI-BLAST
MCL TRIBE
PFAM DB

I am testing these three programs on two sample data sets:

the first is a yeast fillamentus dataset
the second is a general chordate dataset

New Journal Blog

Tue, 24 Jan 2006 14:04:45 GMT

This Journal is ment for bioinformatics and work related posts.

A place for Ideas, Notes and Information.