parameter.space

Q: What are your goals when setting out to build a new product?

A: Our goals are very simple — to design and make better products. If we can’t make something that is better, we won’t do it.

Q: Why has Apple’s competition struggled to do that?

A: Most of our competitors are interested in doing something different, or want to appear new — I think those are completely the wrong goals. A product has to be genuinely better. This requires real discipline, and that’s what drives us — a sincere, genuine appetite to do something that is better.

The other potential scenario, though, is far less optimistic for Apple’s competitors. It’s the iPod model. In this story, Apple begins by releasing a novel, category-defining product. Then, as rivals scramble for some way to respond, Apple relentlessly puts out slightly better versions every year, each time remaining just out of reach of the competition. Meanwhile it lowers its prices and expands its product lineup, making its devices more accessible to a wider audience. Then, to finish the game, it finds a way to boost its position through network effects and customer lock-in. (In the iPod’s case, it accomplished this through the iTunes software and built-in music store.) Put it all together and you have a device that’s unbeatable.

It's becoming harder and harder to see how this isn't the most likely outcome. The only other tablets to move at any volume are being sold AT A LOSS. That's not going to fly, not with the margins Amazon and the like hold in the content world.

This post reminded me of my time wrestling with dissertation project ideas, and the one idea that had me spellbound, but was totally unfeasible. Perturb every known miRNA (up or down, drawbacks aplenty either way) and generate full mRNA and miRNA profiles for each condition.

Putting aside the very real issue of off-target effects, the advantage to this over the Califano method would be a far cleaner signal of what modulating this one molecule does to the genetic makeup of a cell.

It's about time someone devised a good method for integrating existing transcriptome and miRNAome collections into a network of functional predicted functional interactions, as well as miR-mRNA target prediction based on sequence and anticorrelation.

Regarding this post, thanks to some feedback from some of the GenePattern team, both scripts are both available as GenePattern Modules at the GParc! Consider yourself integrated.

Had a labmate come by and ask for help with TCGA Level 3 expression data. Its distributed as one file per sample, which when you have several HUNDRED specimens, isn't really the kind of thing you can or want to re-format by hand if you want to make comparisons across samples.

"How Hard Can it Be?!" I said. Oh when will I learn...

Two days and several infinite loops later, I present two dainty perl scripts for:

a) Take n many sample files and corral them into one flat file, probes x samples. (TCGA Merge)

b) Take a tab delimited text file with x probes and y samples and turn it into a GenePattern-friendly .gct file. (RXC to GCT)

  Still the neatest thing I've seen in 2011. Take whole blood from a person, isolate the intact cells, label and then sort into hundreds of discrete populations. Then, map SIGNALING CHANGES INSIDE THE SAME CELLS back onto the population maps. Still trying to wrap my head around the potential for this in drug development, development and blood cancer biology. Single-Cell Mass Cytometry of Differential Immune and Drug Responses Across a Human Hematopoietic Continuum

Sean C. Bendall, 

Erin F. Simonds, 

Peng Qiu, 

El-ad D. Amir, 

Peter O. Krutzik, 

Rachel Finck, 

Robert V. Bruggner,

Rachel Melamed, 

Angelica Trejo, 

Olga I. Ornatsky, 

Robert S. Balderas, 

Sylvia K. Plevritis, 

Karen Sachs, 

Dana Pe’er,

Scott D. Tanner, 

and Garry P. Nolan

Science 6 May 2011: 332 (6030), 687-696. [DOI:10.1126/science.1198704]

Abstract

  Full Text

  Full Text (PDF)