The goal of this project is to compare rates of evolutionary selection of one-to-one orthologs between the briggsae vs. remanei lineages (by comparison to the species brenneri) for genes with varying levels of sex-specific expression.
Background (M. Palopoli):
Cutter (2005) compared genes between briggsae and elegans and (among other things) found an elevated rate of evolution for sperm-specific genes. He suggested several alternative hypotheses that could not be ruled out but might explain this pattern, including: (1) relaxed selection on sperm function in these two hermaphroditic lineages; and (2) antagonistic coevolution between sperm and oocytes and between the sexes in each lineage.
The former should NOT be true for lineages that still have males and females, such as remanei. However, the latter SHOULD BE EVEN MORE true for lineages with males and females, such as remanei. Hence, this suggests the simple idea of comparing brenneri-briggsae divergence against brenneri-remanei divergence for genes with varying levels of sex-specific expression. The relaxed selection hypothesis predicts that brenneri-briggsae divergences will be greater for male-specific genes; the antagonistic coevolution hypothesis predicts that brenneri-remanei divergences will be greater for male-specific genes. I believe that both hypotheses predict that these differences should disappear for genes expressed in both sexes.
What needs to be done?
Need to figure out the degree to which each gene exhibits sex-specific expression (perhaps using the approach used in Albritton et al.). Also, need to make sure that each gene being compared is a one-to-one ortholog between briggsae and remanei (e.g. reciprocal best blast). Also, best to make sure that each gene is also in a consistent location (at least with respect to X-linkage vs. autosomal). Then need to line up each gene between brenneri and remanei on the one hand, and between brenneri and briggsae on the other hand, and compare Ka values. Perhaps use a simple Chi-square test to see whether there is a nonrandom tendency for either type of comparison to result in higher Ka values for genes in different categories (e.g., male-specific vs. expressed in both sexes).
Background from S. Minot:
A parallel goal of this project is to learn how to execute a computational experiment that adheres to a standard of reproducible research. In the traditional biological sciences there is a well-developed set of standards for the steps that must be taken to ensure that an experiment can be reproduced. However, such standards have not been established for computational biology, and often the experiments described in journal articles cannot be attempted or confirmed by independent researchers due to a lack of code and documentation. This lack of reproducibility is a fundamental barrier to the advancement of computational biology, and anyone working in research (of any sort) should hold themselves and others to a standard of reproducibility.
Reproducibility is also a huge benefit for the advancement of research within a single research group. As a research question expands and is taken on by more investigators, the projects that are structured in a consistent, reproducible way are going to be much easier to build upon and develop.
All of the specific practices outlined here are not the only way to set up a robust research project, but they do use a set of resources and practices that are generally accepted by the research community. Over time the specific pieces of software and conventions will surely change, but I am confident that the general approach will be a benefit for anyone who wants to answer questions using computational tools and communicate their findings to others.
We can break this project up into the following sections:
- Getting Started
- Quantify Sex-Specific Expression
- Identify briggsae/remanei/brenneri Orthologs
- Quantify Rates of Selection
- Statistical Test for Differential Selection among Lineages
A short introduction into the computational practices that will make this project easily reproducible. Very briefly, all analyses will be saved and tracked in this github repository. That way you will always know what the state of the code was at any point that analysis was executed. Also, the computational environment and pre- requisites will all be tracked alongside the rest of the code, so all of the versions of the different software packages that were used can be easily recreated. Lastly, all of the actual computation will take place on a node within the Amazon Web Service (AWS). This has the double benefit of providing useful computational resources, as well as providing a standardized computational environment that is accessible to any other researcher in the field.
See Getting_Started.md for more details.
We will download public RNAseq datasets containing sex-specific transcription information. Albritton, et al. (link) will be a good place to start. Those short reads will be mapped against the appropriate reference genomes, and the number of reads matching each gene will indicate the degree to which each gene is expressed in a sex-specific manner. The output of this step is a ratio of sperm-specificity for each gene.
Datasets from the NCBI SRA database can be downloaded programmatically using the SRA Toolkit (link). While there are many short-read aligner algorithms, BWA (link) is a good standard choice.
Analysis of the output of the short read alignments (and all of the other data outputs produced in this project) will be done in either Python (with Numpy/Pandas) or R. My (Sam's) preference is for Python, but R is also acceptable if that is more familiar.
All of the guidance, code, and output of this section will be kept
in the expression/ folder.
Orthologous genes between the different lineages will be identified the so-called "reciprocal best BLAST" approach. Briefly, every gene from one genome is aligned against every gene from another genome, and vice versa. After applying some basic alignment quality filters, Those pairs of genes that are both the top hit for the other are labeled as being orthologs for the purpose of this analysis.
While BLAST is available as a web tool, there is also a standalone
command line tool that is available for Linux in the ncbi-blast+
package.
All of the guidance, code, and output of this section will be kept
in the orthologs/ folder.
Each pair of orthologous genes will be aligned, and the rates of synonymous and non-synonymous mutation will be calculated. PAML (link) is the standard choice for this type of operation. The resulting dN/dS ratio for each gene will be taken as the output of this step.
All of the guidance, code, and output of this section will be kept
in the selection/ folder.
Given the inputs outlined above:
- Rates of sex-specific gene expression for each gene, and
- dN/dS ratio for each gene, as well as the annotations available for each gene, the final step will be to build a statistical model to test whether there is a significantly different rate of selection between the different pairs of lineages, comparing genes that do and do not exhibit sex- specific expression.
All of the guidance, code, and output of this section will be kept
in the stats/ folder.