Case Study: High Throughput Biology

Loading...

Из курса от партнера University of Michigan

Data Science Ethics

41 оценки

University of Michigan

Data Science Ethics

41 оценки

What are the ethical considerations regarding the privacy and control of consumer information and big data, especially in the aftermath of recent large-scale data breaches? This course provides a framework to analyze these concerns as you examine the ethical and privacy implications of collecting and managing big data. Explore the broader impact of the data science field on modern society and the principles of fairness, accountability and transparency as you gain a deeper understanding of the importance of a shared set of ethical values. You will examine the need for voluntary disclosure when leveraging metadata to inform basic algorithms and/or complex artificial intelligence systems while also learning best practices for responsible data management, understanding the significance of the Fair Information Practices Principles Act and the laws concerning the "right to be forgotten." This course will help you answer questions such as who owns data, how do we value privacy, how to receive informed consent and what it means to be fair. Data scientists and anyone beginning to use or expand their use of data will benefit from this course. No particular previous knowledge needed.

Из урока

Algorithmic Fairness

What could be fairer than a data-driven analysis? Surely the dumb computer cannot harbor prejudice or stereotypes. While indeed the analysis technique may be completely neutral, given the assumptions, the model, the training data, and so forth, all of these boundary conditions are set by humans, who may reflect their biases in the analysis result, possibly without even intending to do so. Only recently have people begun to think about how algorithmic decisions can be unfair. Consider this article, published in the New York Times. This module discusses this cutting edge issue.

Algorithmic Fairness10:12

Correct But Misleading Results12:09

Case Study: High Throughput Biology3:38

Case Study: Geopricing2:52

Case Study: Your Safety Is My Lost Income10:10

Познакомьтесь с преподавателями

H.V. Jagadish
Bernard A Galler Collegiate Professor
Electrical Engineering and Computer Science

Traditionally, biologists worked at a bench

and investigated individual biological entities of interest.

Today, we have something called high throughput biology.

And the way this works is that,

on a single gene chip,

you can have thousands of genes,

all of which can be measured at the same time in parallel.

And so what you're going to do,

is take a few samples of one type of tissue.

Let's say healthy tissue and a few samples of,

let's say disease tissue.

And you just measure the level of gene expression in each for every gene.

Comparing the levels of gene expression in the two tissues,

you can look to find genes where there

was a substantial difference in the expression level,

between the healthy tissue and the disease tissue.

And you can then have something to do with that disease process,

and you could investigate that further.

This kind of high throughput biology has been extremely effective in giving us clues

to many disease processes that we didn't understand before.

However, from a statistical perspective,

there are some issues we should think about.

When you want to compare samples from two populations,

the way that you do this usually is by means of a p-test.

And let's say that you take a p value of 0.95,

what this says is,

that there is a 95% chance that

the difference that you're seeing in the expression levels across the two populations,

is something that was not due to chance.

That there is only a 5% probability that it is due to chance.

Now, this is true for any one single gene,

where the expression level is

different in terms of the p-value that we just talked about.

But now we do this for 25,000 genes.

If for any one single gene,

we have a 5% chance that the difference we're seeing is purely to your chance.

Then if you look at 20 genes,

one of them on average is going to have a difference purely due to chance.

And if you look at 20,000 genes,

a thousand of them on average are going to

have a difference in expression purely due to chance.

The actual number of genes that are involved in

the disease process may be only a few dozen.

And so we have a few dozen genes of interest where the expression level is different,

because there is a real biological difference.

And possibly a few hundred,

maybe a thousand genes where the expression levels are different purely due to chance.

In other words, you've got some information there,

but you've also got a lot of noise.

And how to separate this information from this noise requires,

careful downstream analysis as opposed to

taking what one sees as the gospel truth,

simply because it meets some p-value criteria.

Ознакомьтесь с нашим каталогом

Присоединяйтесь бесплатно и получайте персонализированные рекомендации, обновления и предложения.

Coursera делает лучшее в мире образование доступным каждому, предлагая онлайн-курсы от ведущих университетов и организаций.

© Coursera Inc., 2018 Все права защищены.

Загрузить из App Store

Загрузить в Google Play