File formats

Every file sweepLink and its companion scripts read or write. All of them are plain tab-separated text.

Input files

Sample list

Read by sweeplink-vcf2input, not by sweepLink itself. Tab-separated with a header; lines starting with # are ignored. Three named columns, in any order:

# time in generations before present
sample	population	time_in_gen
HG_TP01_0017	Pop1	130
HG_TP01_0072	Pop1	130

sample

Must match the sample name in the VCF.

population

Free-form label. Becomes the population name everywhere downstream.

time_in_gen

Integer. Whether larger means older is set by --time-direction (bp = before present, forward = forward in time).

Meta file (--meta)

Tab-separated, no header, three columns:

time_130BP_Pop1	0	Pop1
time_115BP_Pop1	15	Pop1
time_101BP_Pop1	29	Pop1

Column	Meaning
1	Sample label. Must appear as a column name in the counts file.
2	Time in generations. Forward, so 0 is the oldest.
3	Population name.

One row per (time point, population) combination. Populations may be missing at individual time points — ragged designs are handled.

Allele counts (--counts)

Tab-separated with a header whose first two fields are placeholders:

-	-	time_130BP_Pop1	time_115BP_Pop1	time_101BP_Pop1
1	579	3/4	15/20	5/8
1	740	0/4	0/20	0/8

Column	Meaning
1	Chromosome.
2	Position.
3+	One per sample label in the meta file, in any order.

Each cell is c/n: c copies of the derived allele out of n sampled chromosomes. c > n is an error.

Important

Missing data is written 0/0, not NA. A cell with n = 0 contributes nothing to the likelihood.

Labels file (--labels, plotting only)

Read by sweeplink-plot to annotate peaks. Tab-separated with a header; needs chr, pos and label columns and ignores any others — so a _selected.txt file with a label column added works directly:

chr	pos	label	label_bottom
2	50001	true sweep	2:50001

Column	Meaning
label	Text above the evidence panel, italic and rotated.
label_bottom	Optional. Text beside the marker in the |s| panel, bold. Defaults to the locus’s own chr:pos.

Rows where both labels are empty are skipped; leaving one empty annotates only the other panel.

Configuration and initial values (--config, --initVals)

--config takes a table with the same columns as <out>_config.txt below. --initVals takes three columns — parameter name, value, jump size.

Output files

Every file below is prefixed with whatever --out was given (sweeplink if omitted). Examples use the tutorial’s prefix, doc.

<out>.parameters — the reproducibility record

Every argument the run used, including the ones left at their defaults and the random seed. Two columns, name and value:

Note

This one file is named after --out; if you did not pass --out it is named after the task instead — infer.parameters, simulate.parameters — not after the default output prefix. Every other file below uses the prefix.

counts	doc_alleleCounts.txt
grid_abs_s	0.01,0.02,0.03,0.04,0.05,0.06,0.07,0.08
iterations	100000
linkage	true
log10N.alpha	16.0
...

Feed it straight back to repeat the run exactly:

$ ./sweeplink doc.parameters

This file is also how the companion scripts discover a run’s settings, which is why they need nothing but --prefix.

<out>_config.txt — per-parameter setup

One row per model parameter: which files it was written to, whether it was updated, which proposal kernel it used, and its prior parameters.

name	priorParameters	traceFile	meanVarFile	...	update	propKernel
log10N	16.0,4.0	doc	doc		1	normal
h		doc_h	doc_h		0	randomInteger

The quickest way to confirm what was actually inferred: update is 0 for anything held fixed.

<out>_trace.txt — the MCMC samples

One column per global parameter, one row per retained iteration (every --thinning-th).

kappa_1	kappa_2	nu_1	nu_2	mu_0	m_Pop1_Pop1	log10N_Pop1	N_Pop1	...
0.000130013	0.134935	0.226948	0.000106703	0.452903	1	4.25177	17855.4

Warning

Use the log10N_* column, not N_*. The chain samples log10N, so N is strongly right-skewed and its arithmetic mean sits far out in the upper tail. The right summary is 10^mean(log10N).

<out>_s_trace.txt holds the same for the per-locus selection coefficients. The matching _h_ and _z_ files are not written by default — see which per-locus files are written.

<out>_meanVar.txt — posterior summaries

name	posterior_mean	posterior_variance
kappa_1	7.3429e-05	1.47218e-09

Computed over all iterations, not just the retained ones — so its numbers will not exactly match a mean taken from the thinned trace.

<out>_{s,h,z}_statePosteriors.txt — the main result

For each locus, the posterior probability of every state on the grid. Row names are <param>_<chr>_<pos>_<population>:

name	state_0	state_1	...	state_8	state_9	...	state_17
s_1_579_Pop1	0	0	...	0.50731	0.49064	...	0

Important

The states are grid indices, not values. Map them back with grid_abs_s from the .parameters file. Because the grid is symmetric, with the default 18 states both index 8 and index 9 mean neutral — any sum over states has to add the two together.

Rows are accumulated over every iteration and sum to 1, so the underlying counts recover as k = round(p × iterations).

<out>_{s,h,z}_posteriorMode.txt

The single most probable state per locus — again an index, not a value.

name	posterior_mode
s_1_579_Pop1	8

<out>_LL*.txt — likelihood surfaces

Written by the four calculateLL* tasks, each to its own file — <out>_LL.txt, <out>_LLForS.txt, <out>_LLForSH.txt, <out>_LLForN.txt — so they can share an --out. Columns depend on the task; for calculateLLForN with --sumLL there is a single row:

chr	pos	LL0_N_1.000000	LL1_N_1.149757	...
sumLL	-	-4954.27	-4953.52	...

<out>_selected.txt

Written by sweeplink-extract, not by sweepLink:

chr	pos	direction	mean_s	samples_supporting	samples_against	posterior	run
2	8126	negative	-0.0220707	100000	0	0.99999	doc

mean_s is the signed posterior mean of the selection coefficient; posterior is the smoothed probability that the sign is as reported.