Courses

Course Materials

Bioinformatics:

Python:

Linux:

Máster en mejora genética:

  • Materiales para el curso de Aplicaciones de la genómica a la Mejora vegetal.

  • Materiales para el curso de Aplicaciones de la genética de poblaciones a la Mejora vegetal.

Grado en ingeniería biomédica:

Seminario de agricultura y moral.

Software

Population genetics simulations are the simulations written for our population genetics practices. Each one is a small interactive application: set the parameters of a population, run it, and watch the allele frequencies respond.

Sequence comparison tools teach basic bioinformatic concepts: the dotplot, inversions, manual alignment, which scores an alignment typed by hand, Smith-Waterman, and BLAST words, which turn a database search into a lookup.

SNP calling is the practice for statistical inference: one simulated set of reads called three times over – with thresholds, by maximum likelihood and by Bayes – and scored against a truth that is known because it was simulated.

Do cows fly? is the p-value practice. A friend claims that cows fly, for so many hours a day. Every visit to every farm finds every cow on the ground, so what changes from one experiment to the next is the farm – and a field of nine cows with nothing in the air is the same observation as a field of ninety, and nothing like the same evidence. At one hour a day it takes a farm of 71 cows before that empty sky rejects the claim.

A positive HIV test is the inverse probability practice. The doctor can quote the test’s false positive rate, P(positive | not infected); what the person wants is P(infected | positive). The page works both out from the prevalence and the error rates of the test, and shows the same answer again as whole people.

The likelihood of a coin is the maximum likelihood practice, done by hand. The first tab works out the likelihood of one P(heads) – the probability of the count of heads seen, if the coin landed heads that often – and the second moves P(heads) with a slider, leaving a trail, to go looking for the value that makes the tosses likeliest.

A coin, three ways takes one string of coin tosses and reasons about it three times over: a classical test, maximum likelihood and Bayes, with all three answers on show at once. Ten heads out of ten reject a fair coin, estimate the bias as exactly one with an interval of zero width, and leave a posterior mean of 0.917: the same tosses, three languages. Toss by toss, one at a time, so how much evidence it takes to move each of them can be seen rather than asserted.

All of them run entirely in the browser: there is nothing to install, and no sequence you type is sent anywhere.