HMM¶
The HMM method can be used to determine the essentiality of the entire genome, as opposed to gene-level analysis of the other methods. It is capable of identifying regions that have unusually high or unusually low read counts (i.e. growth advantage or growth defect regions), in addition to the more common categories of essential and non-essential.
Note
Intended only for Himar1 datasets.
How does it work?¶
For a formal description of how this method works, see our paper [DeJesus2013HMM]:
DeJesus, M.A., Ioerger, T.R. A Hidden Markov Model for identifying essential and growth-defect regions in bacterial genomes from transposon insertion sequencing data. BMC Bioinformatics. 2013. 14:303
Example¶
python3 transit.py hmm <comma-separated .wig files> <annotation .prot_table or GFF3> <output file>
Optional Arguments:
-r <string> := How to handle replicates. Sum, Mean. Default: -r Mean
-l := Perform LOESS Correction; Helps remove possible genomic position bias. Default: Off.
-iN <float> := Ignore TAs occuring at given percentage (as integer) of the N terminus. Default: -iN 0
-iC <float> := Ignore TAs occuring at given percentage (as integer) of the C terminus. Default: -iC 0
Parameters¶
The HMM method automatically estimates the necessary statistical parameters from the datasets. You can change how the method handles replicate datasets:
- Replicates: Determines how the HMM deals with replicate datasets by either averaging the read-counts or summing read counts across datasets. For regular datasets (i.e. mean-read count > 100) the recommended setting is to average read-counts together. For sparse datasets, it summing read-counts may produce more accurate results.
Output and Diagnostics¶
The HMM method outputs two files. The first file provides the most
likely assignment of states for all the TA sites in the genome. Sites
can belong to one of the following states: “E” (Essential), “GD”
(Growth-Defect), “NE” (Non-Essential), or “GA” (Growth-Advantage). In
addition, the output includes the probability of the particular site
belonging to the given state. The columns of this file are defined as
follows:
| Column # | Column Definition |
|---|---|
| 1 | Coordinate of TA site |
| 2 | Observed Read Counts |
| 3 | Probability for ES state |
| 4 | Probability for GD state |
| 5 | Probability for NE state |
| 6 | Probability for GA state |
| 7 | State Classification (ES = Essential, GD = Growth Defect, NE = Non-Essential, GA = Growth-Defect) |
| 8 | Gene(s) that share(s) the TA site. |
The second file provides a gene-level classification for all the
genes in the genome. Genes are classified as “E” (Essential), “GD”
(Growth-Defect), “NE” (Non-Essential), or “GA” (Growth-Advantage)
depending on the number of sites within the gene that belong to those
states.
| Column Header | Column Definition |
|---|---|
| Orf | Gene ID |
| Name | Gene Name |
| Desc | Gene Description |
| N | Number of TA sites |
| n0 | Number of sites labeled ES (Essential) |
| n1 | Number of sites labeled GD (Growth-Defect) |
| n2 | Number of sites labeled NE (Non-Essential) |
| n3 | Number of sites labeled GA (Growth-Advantage) |
| Avg. Insertions | Mean insertion rate within the gene |
| Avg. Reads | Mean read count within the gene |
| State Call | State Classification (ES = Essential, GD = Growth Defect, NE = Non-Essential, GA = Growth-Defect) |
Note: Libraries that are too sparse (e.g. < 30%) or which contain
very low read-counts may be problematic for the HMM method, causing it
to label too many Growth-Defect genes.
Run-time¶
The HMM method takes less than 10 minutes to complete. The parameters
of the method should not affect the running-time.