There is a weird lag at the moment when loading a page, especially where the page contains a table, where the initial table style is different then I assume some CSS resource gets loaded and the table changes. Maybe due to bokeh js and css getting loaded before bootstrap?

Currently the maximum percentile for each signal is computed as the percentile of the maximum statistic value within the peak against values for all windows. It would be more informative and appropriate to communicating the outlier status of a peak if percentiles were computed against the raw (i.e., per-SNP) statistic values.

IHS peaks can be very broad and/or irregular, which can lead to adjacent peaks being merged or incorrect targeting of known loci.

Here's a case where Gste is not targeted: https://malariagen.github.io/agam-selection-atlas/0.1-alpha3/signal/IHS/BFS/3/1/index.html

Here's a case where two peaks are merged: https://malariagen.github.io/agam-selection-atlas/0.1-alpha3/signal/IHS/BFS/X/1/

Consider adding some interactive scatter plots to the site to provide an overview of signals, where each signal is plotted as a point on a scatter plot, with the peak model fit score as the X axis and the max percentile as the Y axis. Signals with strong evidence on both these scores should then appear in the top right of the plot. If the plot was done with bokeh, hover could be added to each signal similar to the hover currently used in the signal summary plots over the genome, including a preview of the peak data and fit. Also clicking on a signal could link through to the signal page.

• Add labelled genes track
• (Add heterochromatin track)

Produce some kind of overview plot showing a whole chromosome (arm) with all signals marked. Possibly use colours for populations and marker shapes for statistics. Even better if could be to some extent interactive, i.e., can hover over signals and follow links to signal pages.

This would be good for the all signals page as well as for the population pages.

Add a tutorial section with a walk-through of how to use the site.

E.g., make them bold.

Consider writing an executive summary, as a page of prose with brief explanation plus hand-written summary of the major selection signals we think are important and which populations and genes they involve, for anyone coming to the site who isn't a genetics geek.

The site currently has GNS scans for H12 and IHS, but there were no GNS comparison in the raw XPEHH set. If we want to be able to compare the number of signals per population including Guinea, we should add XPEHH for GNS (probably using same comparisons as BFS).

Add content for all figure captions.

Add content to methods section.

• Data sources (i.e., what data were used, where did they come from)
• IHS
• XPEHH
• H12
• XPCLR
• Delta Tajima's D
• Peak finding algorithm

Consider adding pages and interactive plots to allow browsing of the windowed data for each selection statistic and population, plotted over whole chromosome arms. This would allow unconstrained browsing of the data, for those who want to see what else may be there that is not getting picked up by the peak finding algorithm.

See https://www.biorxiv.org/content/biorxiv/early/2017/12/29/240747.full.pdf for information about recent reannotation of An. gambiae genome and discovery of 64 new IRs.

Investigate use of data tables for selection signals, so user can change the sort order (and possibly also create filters/queries).

Figure out how to include versioning into the site.