In gthar/NucleosomeDynamics: Nucleosome Dynamics

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Nucleosome Dynamics allows detection of changes in the positions of nucleosomes at single read level between two reference nucleosome maps. Changes include shifts (moves), indels (insertions/deletions) of the existing reads in initial state to match the reference final state.

Method

Nucleosome Dynamics is a method that allows comparison of two diferent MNase-seq experiments (Condition 1 and Condition 2) and detects local changes like:

• Upstream shifts, were the occupancy is shifted upstream on Condition 2 relative to Condition 1.
• Downstream shifts, were the occupancy is shifted downstream on Condition 2 relative to Condition 1.
• Inclusions, where the nucleosome occupancy is significantly increased.
• Evictions, where the nucleosome occupancy is significantly decreased.

Nucleosome Dynamics tries to identify those changes by working at a read-level. The way its pipeline works is by sequentially pairing reads of one experiment with reads of the second experiment and setting those paired reads apart.

First, all reads that are identical on both experiments are separated. Then, the reads that either start or end at the same position are separated as well. Next, the reads from one experiment whose range is completely contained by the range of a read in the other experiment, are paired and separated too. This first pairing, account for either differences in MNase digestion or for no differences at all, so after removing those, what's left accounts for differences.

Nucleosome Dynamics then pairs reads from one experiment to reads in the other experiment whose center is slightly shifted either upstream or downstream relative to each other. In an attempt to find the best possible combination of pairs, a dynamic programming algorithm is used. The dynamic programming algorithm is set so that:

1. The most number of pairs possible is found.
2. The centers of the paired reads are as close as possible.
3. Reads whose center is futher apart than 74 bp. are never paired.

To achieve that, the dynamic programming algorithm works the following way:

1. Gaps are highly penalized (to achieve maximum number of pairs).
2. Pairs are given a score that is inversely proportional to the centers distance (to prioratize close pairs).
3. Centers distance bigger than 74 bp. are given a score of -Infinity so that it can never happen.

Once the shift pairs have been found, accumulations of such are considered shift hotspots.

In order to identify significant changes in occupancy (insertions and deletions), the coverage of the reads in both experiments is used. First a normalized z score accross the genome is calculated, assuming a hypergeometric distrubition. This score is normalized on a 10000 bp. windows, so it takes into account fluctuations of coverage accross big segments of the genome and fins locally significant differences.

Positive peaks of that z score mean that the coverage of the Experiment 1 is significantly higher than the coverage of the Experiment 2, and therefore it's considered a deletion. Similarly, negative z score peaks represent regions where the coverage of the Experiment 2 is significantly higher and therefore classified as insertions.

Finally, the Nucleosome Dynamics pipeline scores the hotspots found. To do so, a p-value is calculated for each point on the sequence, that accounts statistically significant differences of coverage on a 10000 bp. window. On 10000 bp. windows and for each position, a Fisher's test is calculated with the following contingency table.

Each hotspot is given the p-value at its peak position as score. Then a threshold is applied so that only the more relevant hotspots are reported.

gthar/NucleosomeDynamics documentation built on May 17, 2019, 8:56 a.m.