Lessons from today's presentation

1. Don't speak too fast. Make yourself clear.
2. Case study is very important to demonstrate the usefulness of your program, especially if you want to convince your audience.
3. Explain the concepts clear before going to your results, otherwise the audience would be confused and lost in the presentation
4. For any anomalies in the results, better to investigate the cause. They are popular places where people would ask question.
5. Explain the particular reason for choosing, e.g. hierarchical clustering. Why not other clustering methods? (The reason is that k-means and density-based clustering require spatial points; I've only got the distance matrix.
6. Demonstrate improvements over past work.

Datamining routines

Routines are:

1. Fill in blank cells (process missing data)
2. Feature selection
3. Remove outliers
4. Train classifier
5. Validate the classifier trained

After or before doing the step 2, one may want to discretize the features to make them categorical.

Each of the step may have many proposed algorithms to complete. One probably has to try many of them to get a high accuracy classifier.

Archving NUS Web mail in Thunderbird

0. Create a fold "NUS Emails Archive" under "Local Folders"

My folder structure in Thunderbird become:
Local Folders
->Unsent
->Trash
->NUS Emails Archive
NUS Email
->Inbox
->Drafts
->Sent

1. Click Inbox under NUS Email
2. Drag all emails you want to archive locally to NUS Email Archive
3. Go to "C:\Documents and Settings\user name\Application Data\Thunderbird\Profiles\2r1eew7w.default\Mail\Local Folders" to view the file "NUS Emails Archive" and its size change.

Rediscovery of the power of R

R is really nice and powerful in producing graphs. A lot to learn!

Google Help : Cheat Sheet

From http://www.google.com/help/cheatsheet.html

Here's a quick list of some of our most popular tools to help refine and improve your search. For additional help with Google Web Search or any other Google product, you can visit our main Google Help page.
OPERATOR EXAMPLE		FINDS PAGES CONTAINING...
vacation hawaii		the words vacation and Hawaii .
Maui OR Hawaii		either the word Maui or the word Hawaii
"To each his own"		the exact phrase to each his own
virus –computer		the word virus but NOT the word computer
+sock		Only the word sock, and not the plural or any tenses or synonyms
~auto loan		loan info for both the word auto and its synonyms: truck, car, etc.
define:computer		definitions of the word computer from around the Web.
red * blue		the words red and blue separated by one or more words.
I'm Feeling Lucky		Takes you directly to first web page returned for your query.
CALCULATOR OPERATORS		MEANING		TYPE INTO SEARCH BOX
+		addition		45 + 39
-		subtraction		45 – 39
*		multiplication		45 * 39
/		division		45 / 39
% of		percentage of		45% of 39
^		raise to a power		2^5 (2 to the 5th power)
ADVANCED OPERATORS		MEANING		WHAT TO TYPE INTO SEARCH BOX (& DESCRIPTION OF RESULTS)
site:		Search only one website		admission site:www.stanford.edu (Search Stanford Univ. site for admissions info.)
[#]…[#]		Search within a range of numbers		DVD player $100..150 (Search for DVD players between $100 and $150)
link:		linked pages		link:www.stanford.edu (Find pages that link to the Stanford University website.)
info:		Info about a page		info:www.stanford.edu (Find information about the Stanford University website.)
related:		Related pages		related:www.stanford.edu (Find websites related to the Stanford University website.)

©2008 Google

Medicel

"Medicel develops technologies and methods to help scientists turn data into information and knowledge." A sophisticated platform in which data from various sources e.g. HPLC, MS can be processed. It would be wonderful if they also have a data mining feature to run commonly used machine learning and data mining algorithms. From a developer's perspective, this platform is excellent, because it satisfies various daily requirements. It is great with 100 engineers developing 7 years. With moderate training, a junior engineer without much biological background can process tremendous amount of data from biological and medical experiments in various formats very efficiently. The cost of the software is also very high --- half a million.

Doctors do not like this software, though, for reasons that I can understand. From a doctor's perspective, it is good enough to have a small tool to complete a specific task. Simplicity is beauty. Too many functions and features are too distracting. They would think I am a doctor, not a programmer. Give me minimum advice, and I can get the work done.

It is still a valuable tool for a bioinformatician to do case analysis. Some ideas in that software can benefit us in our future plan of designing a similar but far simpler platform.

Future projects and project management tools

1. Design biological platform so that softwares and program in analyzing gene expression etc. can be integrated. (propose to James)
2. Outlier analysis (Try to find its application in biological context; customer relation management is a potential application area )
3. Program that accepts the IC values and outputs the desired cutting point so that gene selection achieves certain accuracy
4. Imaging processing tool to analyse the cancer cell lines
5. Android (pending)

Potential project management portals are: SourceForge.net and Assembla. wetpaint is a nice project management wiki, but Assembla provides both wiki and subversion, which are more suitable for team projects.