Nucleosome Occupancy Likelihood (NOL Plots) Predicted for the Maize Genome
Support Vector Machines (SVMs) are supervised learning classifiers that can be used to predict the nucleosome forming or nucleosome inhibiting potential for any given DNA sequence. Steps to train a model to discriminate between nucleosome forming and nucleosome inhibiting sequences are:
1. Expeimentally identify the most nucleosome forming and most nucleosome inhibitory sequences
2. Score sequences based on their abundance in either class.
3. Train a support vector machine to discriminate between the two data sets.
4. Assign a score to each 50-mer of the maize genome (base by base) for nucleosome forming or inhibitory potential.
Previously, we have measured nucleosome positions at 425 human transcription start sites and used these data to train a human-based SVM to identify the nucleosome forming potential (nucleosome occupancy likelihood, NOL) of any human DNA sequence. We applied the human-trained SVM to the maize genome and have identified similar patterns of nucleosome distributions as human nucleosome distributions at TSSs. The human-trained SVM also reveals striking features of predicted nucleosome position at other genetic elements in maize. To investigate the basis and evolutionary conservation of sequence-driven nucleosome distribution in maize, we are training a maize-based support vector machine by experimentally measuring nucleosome distribution at 400 transcription start sites.
Click here to view the human-based NOL applied to the maize genome in the MaizeGDB genome browser.
New Publication on NOL plots and Genomaize Browser for Maize B73
(2013)
Fincher JA, Vera DL, Hughes DD, McGinnis KM, Dennis JH, and Bass HW.
"Genome-wide prediction of nucleosome occupancy in maize (Zea mays L.) highlights chromatin structural features at multiple scales." Plant Physiology
(in press)
|