Fitness models out of SATAY data

[leilaicruz]

Lets put here different ways you have tried out to get fitness from satay data

[leilaicruz]

Fitness from an intergenic competition model

This model describes a population fitness that obeys a logistic expression due to the scarcity of resources on the environment that effectively sets a limit on the growth rate of the entire population.

Here, one of the assumptions will be that :

• at the start of the reseed (population growth phase) the number of reads is one for all the transposon insertions , so no growth has hapened yet during the induction part. Effectively we have a single copy of each genotype in the population.
• The solution for the growth rates will be : np.log(N/(1-N/K))/T where N is the N=['number_of_read_per_gene']/(['number_of_transposon_per_gene']-1) , K is the so called carrying capacity , which is the maximum N the population can achieve , in this case will be the total_reads/total_tn from the population , and T is the time interval of this process that will be 90 hours (reseeding time). Look up here for the solution of the logistic expression
• The code for this is available in : https://github.com/leilaicruz/python-modules-for-bioinformatic-analyses/blob/project_code/src/script_fitness-from-reads-per-transposon-satay-data.py

Output from this approach

• With this approach you can generate the same kind of volcano plots to pinpoint significant genes that has a high log fold change and high negative p-value among two libraries. In the figure below I show the same plot for the fitnesses values (growth rates out of the intergenic competition model) and from the reads per transposons values.

  Some testing on existing measurements

• When plotting the logistic growth with the estimated growth rate for a specific gene in that time interval , we can see for example in the case of Bem1 how in the dnrp1 can manage to grow from the 60 hours in the reseeding while for the WT it can not. This is somehow what we expect for the case of Bem1. The same can be plotted for Bem3 that we know that in dnrp1 it is a bit slower than in the WT.
  

Remaining questions

• Which normalization to N should I still use?
• Is this definition of K reasonable?
• Other comments