Presentation is loading. Please wait.

Presentation is loading. Please wait.

False discovery rate estimation

Similar presentations


Presentation on theme: "False discovery rate estimation"— Presentation transcript:

1 False discovery rate estimation
Genome 559: Introduction to Statistical and Computational Genomics Prof. William Stafford Noble

2 One-minute responses Overall I feel I’m getting the hang of it.
Great sequence of problem solving. I like the questions that build on previous code. Today’s class was helpful in terms of learning how to break down problems. Relevant programming makes me happy. Good pace. I liked the integration of the programming with concepts.

3 One-minute responses: questions and requests
A bit fast. x3 Not enough time for any of the practice problems today. Today’s problem was very challenging. I tend to grasp the math parts of lecture after class. Are there any resources you could recommend for stats & comp bio. I added a pointer for today’s content. You can also look at the “Points of Significance” column in Nature Methods ( tatistics_the_attention_it_deserves.html) Please go over the math stuff again. Still having tricky time looping over matrices. Screenshots of the text editor would be more helpful than using text in powerpoint for sample problems. Can we spend more time on the motif practice problem next class? What will the exam look like? Similar to homeworks, except the programming will be on paper and will be short questions. Open book. For the EVD, do the distributions always look the same given that the DP matrix can be different? No, the distribution for a given motif will not always precisely follow an EVD. And it does differ from one motif to the next. It would be helpful if we could see the DP for motifs written out to the solution. Can we go over how to fill the histogram in class?

4 Can you discuss conditionals we add to for loops like and, or
Can you discuss conditionals we add to for loops like and, or? Whether order of conditionals matters.

5 Converting scores to p-values
A C G T 1 1 1 1 1 In the end, the bottom row contains the scores for all possible sequences of length N. Use these scores to compute a p-value.

6

7

8

9

10 Cumulative distribution of counts in the final row of the DP matrix
Number of sequences Motif score

11

12

13 False discovery rate: Motivation
Scenario #1: You have used PSI-BLAST to identify a new protein homology, and you plan to publish a paper describing this result. Scenario #2: You have used PSI-BLAST to discover many potential homologs of a single query protein, and you plan to carry out a wet lab experiment to validate your findings. The experiment can be done in parallel on 96 proteins.

14 Types of errors False positive: the algorithm indicates that the sequences are homologs, but actually they are not. False negative: the sequences are homologs, but the algorithm indicates that they are not. Both types of errors are defined relative to some confidence threshold. Typically, researchers are more concerned about false positives.

15 False discovery rate 5 FP 13 TP 33 TN 5 FN
The false discovery rate (FDR) is the expected percentage of target sequences above the threshold that are false positives. In the context of sequence database searching, the false discovery rate is the percentage of sequences above the threshold that are not homologous to the query. 33 TN 5 FN Homolog of the query sequence Non-homolog of the query sequence FDR* = FP / (FP + TP) = 5/18 = 27.8% *Technically, this is the false discovery proportion, and the FDR is the expectation of the FDP.

16 Bonferroni vs. FDR Bonferroni controls the family-wise error rate; i.e., the probability of at least one false positive among the sequences that score better than the threshold. The Benjamini-Hochberg procedure controls the false discovery rate; i.e., the expected percentage of false positives among the target sequences that score better than the threshold.

17 Controlling the FDR Order the unadjusted p-values p1  p2  …  pm.
To control FDR at level α, Reject the null hypothesis for j = 1, …, j*. (Benjamini & Hochberg, 1995)

18 Benjamini-Hochberg example
Rank (jα)/m p-value Choose the largest threshold j so that (jα)/m is less than the corresponding p-value. Approximately 5% of the examples above the line are expected to be false positives.

19 Benjamini-Hochberg test
Test of 100 uniformly distributed p-values (p- values from non-significant results) P-values as blue dots Significance threshold for FDR = 0.2 as red line An idealized experiment in which 100 cases, none of which are significant, are tested with the Benjamini-Hochberg procedure, controlling the false discovery rate at 20%. The blue dots are the ranked p-values from the 100 cases, and the red line is the significance threshold established by the Benjamini-Hochberg procedure. None of the cases can be declared significant.

20 Benjamini-Hochberg test
Test of 10 low p-values (significant results) mixed with 90 p-values from non- significant results P-values as blue dots Significance threshold for FDR = 0.2 as red line Eleven cases declared significant Declare significant An idealized experiment in which 10 cases with significantly low p-values are mixed with 90 cases that are not significant. All cases can be declared significant up to the highest-ranked case that falls below the significance threshold.

21 Summary Selecting a significance threshold requires evaluating the cost of making a mistake. Bonferroni correction divides the desired p-value threshold by the number of statistical tests performed. The E-value is the expected number of times that the given score would appear in a random database of the given size. The false discovery rate is the percentage of false positives among the target sequences that score better than the threshold. Use Bonferroni correction when you want to avoid making a single mistake; use Benjamini-Hochberg when you can tolerate a certain percentage of mistakes.

22 Hint: <list>.sort() will sort your list.
Sample problem #1 Rank (jα)/m p-value Choose the largest threshold j so that (jα)/m is less than the corresponding p-value. Given: a confidence threshold, and a list of p-values Return: a set of p-values with the specified false discovery rate ./compute-fdr.py 0.1 pvalues.txt Hint: <list>.sort() will sort your list.

23

24

25 Sample problem #2 Modify your program so that it will work with an arbitrarily large collection of p-values. You may assume that the p-values are given in sorted order. Read the file twice: once to find out how many p-values there are, and a second time to do the actual calculation.


Download ppt "False discovery rate estimation"

Similar presentations


Ads by Google