Make Chainfile:

Clone With Submodule

Clone this repository and initialize AnchorWave/ in one step:

git clone --recurse-submodules <repo-url>

If you already cloned without submodules:

git submodule update --init --recursive

Build AnchorWave (submodule)

Build the AnchorWave binary from the included submodule:

cd AnchorWave
cmake ./
make -j
cd ..

The alignment script will use anchorwave from your PATH if available, or fall back to the local binary under AnchorWave/.

Files

v2 and v5 maize genome files, from maizegdb:

• v2.fa.gz
• v5.fa.gz

You also need a reference annotation file (.gff3 or .gtf) matching the reference genome used with --ref.

• v5.gff.gz

Scripts:

• scripts/align_with_anchorwave.sh

Lift-Over to B73 v5

To move marker coordinates from AGPv2 to B73 v5, we converted the resulting MAF alignment from Anchorwave into a UCSC chain file with maf-convert:

maf-convert chain AW_out/alignment.maf > v2v5.chain

Requirements

Run module load last and module load minimap2 (or install both tools so they are available on PATH).

Quick Start

Run with explicit inputs:

scripts/align_with_anchorwave.sh \
  --gff path/to/reference.gff3 \
  --ref B73_RefGen_v2.fa.gz \
  --query Zm-B73-REFERENCE-NAM-5.0.fa.gz \
  --threads 16 \
  --outdir anchorwave_out

Typical run from repository root after building AnchorWave:

scripts/align_with_anchorwave.sh \
  --gff v5.gff3.gz \
  --ref v5.fa.gz \
  --query v2.fa.gz \
  --threads 16 \
  --outdir anchorwave_out

Output

By default (--outdir anchorwave_out):

• anchorwave_out/alignment.maf
• anchorwave_out/alignment.fragments.maf
• anchorwave_out/anchors.txt
• anchorwave_out/anchorwave.log

Notes for Maize-Scale Genomes

• Start with --threads 16 (or higher based on available CPU).
• Use SSD-backed storage for faster SAM/MAF I/O.
• Keep sufficient free disk space: intermediate and final alignment files can be large.
• Default mode is proali; switch to genoali with --mode genoali if needed.

Script Help

scripts/align_with_anchorwave.sh --help

Ogut Map

OGUT Marker Map Provenance

The OGUT marker map in this repository traces back to the marker resource reported in:

• Ogut F, Bian Y, Bradbury PJ, Holland JB. 2015. Joint-multiple family linkage analysis predicts within-family variation better than single-family analysis of the maize nested association mapping population. Heredity. https://doi.org/10.1038/hdy.2014.123

The original marker table was recorded here as ogut_fifthcM_map_agpv2.csv, then converted to BED format as ogut_fifthcM_map_agpv2.bed. The BED conversion uses chromosome, physical position, marker name, source SNP ID, and cM value, with 0-based half-open coordinates:

awk -F',' 'NR>1 { pos=$4+0; print $3 "\t" pos-1 "\t" pos "\t" $2 "\t" $1 "\t" $5 }' \
  ogut_fifthcM_map_agpv2.csv > ogut_fifthcM_map_agpv2.bed

Move to v5

After building the chain file, the marker BED was translated to v5 coordinates with CrossMap:

CrossMap bed v2v5.chain ogut_fifthcM_map_agpv2.bed ogut_v5.bed

This produced ogut_v5.bed, which contains the lifted marker positions on the v5 assembly while preserving the marker IDs and cM annotations.

Post-Lift Manual Marker Adjustments

After inspecting ogut_v5.bed, we found markers whose physical coordinates were not monotonically increasing within chromosome even though the cM values remained monotonic. To keep the map order consistent with the physical order on v5, we adjusted the cM values for the affected markers without changing their lifted physical positions.

Applied adjustments:

• On Chr3, M2179 and M2180 kept their physical coordinates, but their cM values were swapped.
• On Chr7, the block from M5158 through M5171 kept its physical coordinates, and the cM values across that block were reversed so that genetic order matches the physical order.

The pre-edit ogut_v5.bed was committed as ce1ab1f, and a local backup copy was also written to ogut_v5.bed.bak before those edits.

The final map is visualized in ogut_v5_cm_vs_bp_by_chr.png.

Rodgers-Melnick Map

Rodgers-Melnick Crossover Map Provenance

The Rodgers-Melnick crossover intervals in this repository trace back to:

• Rodgers-Melnick E, Bradbury PJ, Elshire RJ, Glaubitz JC, Acharya CB, Mitchell SE, Li C, Li Y, Buckler ES. 2015. Recombination in diverse maize is stable, predictable, and associated with genetic load. PNAS. https://doi.org/10.1073/pnas.141386411

The original interval tables were downloaded from Panzea and recorded here as:

• RodgersMelnick2015PNAS_cnnamImputedXOsegments.txt
• RodgersMelnick2015PNAS_usnamImputedXOsegments.txt

Build the v2 BED with Unique IDs

We first removed heterozygous intervals, extracted chromosome/start/end plus family, and appended a unique ID to each v2 interval:

cat RodgersMelnick2015PNAS_cnnamImputedXOsegments.txt RodgersMelnick2015PNAS_usnamImputedXOsegments.txt | \
  grep -v 'het' | grep -v 'Family' | \
  tee >(cut -f 1,5,6 > left.txt) | cut -f 2 > right.txt

paste left.txt right.txt | \
  sed -e 's/\r//g' | \
  awk 'BEGIN{OFS="\t"} {print $0, sprintf("RMv2_%06d", NR)}' \
  > RodgersMelnickv2.bed

This produced RodgersMelnickv2.bed with columns:

• chromosome
• start
• end
• family
• unique interval ID

Lift to v5

After building the chain file, the v2 interval BED was translated to B73 v5 coordinates with CrossMap:

CrossMap bed v2v5.chain RodgersMelnickv2.bed RodgersMelnickv5.bed

Because these are long intervals rather than single markers, many v2 regions were split into multiple lifted fragments in v5.

Merge Split Fragments and Remove Multi-Fragment Regions

To reconnect nearby lifted fragments from the same original v2 interval and family, we merged v5 fragments when they were on the same chromosome, had the same family and interval ID, and were within 100 kb:

sort -k4,4 -k5,5 -k1,1V -k2,2n RodgersMelnickv5.bed | \
awk 'BEGIN{OFS="\t"}
NR==1{chr=$1; s=$2; e=$3; fam=$4; id=$5; next}
$1==chr && $4==fam && $5==id && $2<=e+100000 {if($3>e)e=$3; next}
{print chr,s,e,fam,id; chr=$1; s=$2; e=$3; fam=$4; id=$5}
END{if(NR) print chr,s,e,fam,id}' > RMv5_combined.bed

We then removed any original v2 interval IDs that still appeared more than once after this merge, keeping only uniquely resolved lifted regions:

awk 'BEGIN{FS=OFS="\t"} {count[$5]++; line[NR]=$0; id[NR]=$5} END{for(i=1;i<=NR;i++) if(count[id[i]]==1) print line[i]}' \
  RMv5_combined.bed > RMv5_unique.bed

Convert Unique Regions into a Summed Rate Track

Each interval in RMv5_unique.bed represents one crossover localized to a region. To build a per-base crossover-rate track, we assigned each interval a uniform weight of 1 / interval_length, emitted start and stop events, sorted them, and then swept across each chromosome to sum the active rates:

awk 'BEGIN{OFS="\t"}
{
  len = $3 - $2 + 1
  rate = 1 / len
  print $1, $2, rate
  print $1, $3 + 1, -rate
}' RMv5_unique.bed > RMv5.events.tsv

sort -k1,1V -k2,2n RMv5.events.tsv > RMv5.events.sorted.tsv

awk 'BEGIN{OFS="\t"; eps=1e-12}
function clean(x) {
  return (x < eps && x > -eps) ? 0 : x
}
NR==1 {
  chr = $1
  pos = $2
  delta = $3 + 0
  cur = 0
  next
}
$1 == chr && $2 == pos {
  delta += $3
  next
}
{
  cur = clean(cur + delta)

  if ($1 == chr && cur != 0 && pos < $2) {
    print chr, pos, $2 - 1, cur
  }

  if ($1 != chr) {
    cur = 0
  }

  chr = $1
  pos = $2
  delta = $3 + 0
}
END{
  cur = clean(cur + delta)
}' RMv5.events.sorted.tsv > RMv5_rates.bed

This sweep collapses all event deltas at the same position, applies the total change once, and then emits the constant-rate interval to the next event position. The final file, RMv5_rates.bed, is a chromosome-wise interval track in which each region stores the summed crossover rate over all uniquely resolved Rodgers-Melnick intervals lifted to B73 v5.

Summarize Region CO and Chromosome cM Lengths

We next converted each RMv5_rates.bed interval into its total crossover contribution across the interval:

awk 'BEGIN{FS=OFS="\t"} {print $0, $4 * ($3 - $2 + 1)}' RMv5_rates.bed > RMv5_CO.bed

To anchor the Rodgers-Melnick map to the chromosome-scale genetic lengths in the Ogut v5 map, we computed chromosome cM spans from ogut_v5.bed:

awk 'BEGIN{FS=OFS="\t"}
{
  cm = $6 + 0
  if (!($1 in min) || cm < min[$1]) min[$1] = cm
  if (!($1 in max) || cm > max[$1]) max[$1] = cm
  n[$1]++
}
END{
  print "chromosome","cM_length","min_cM","max_cM","n_markers"
  for (chr in n) {
    print chr, max[chr] - min[chr], min[chr], max[chr], n[chr]
  }
}' ogut_v5.bed | sort -k1,1V > ogut_v5_chromosome_cM_lengths.tsv

Rescale Rodgers-Melnick Regions to Ogut cM Lengths

Finally, we rescaled the Rodgers-Melnick interval contributions so that the summed crossover signal on each chromosome matches the total chromosome cM length from ogut_v5_chromosome_cM_lengths.tsv. We also computed cM/Mb for each region:

awk 'BEGIN{FS=OFS="\t"}
NR==FNR && FNR>1 { chr_cm[$1] = $2; next }
{
  len = $3 - $2 + 1
  val = $4 * len
  chr_sum[$1] += val
  row[NR] = $0
  row_val[NR] = val
  row_len[NR] = len
}
END{
  print "chr","start","end","rate","scaled_cM","cM_per_Mb"
  for (i = 1; i <= NR; i++) {
    split(row[i], f, FS)
    scaled_cm = (chr_sum[f[1]] > 0 ? row_val[i] * chr_cm[f[1]] / chr_sum[f[1]] : 0)
    cm_per_mb = (row_len[i] > 0 ? scaled_cm / row_len[i] * 1e6 : 0)
    print f[1], f[2], f[3], f[4], scaled_cm, cm_per_mb
  }
}' ogut_v5_chromosome_cM_lengths.tsv RMv5_rates.bed > RMv5.bed

This final RMv5.bed contains:

• chromosome
• start
• end
• summed Rodgers-Melnick crossover rate
• chromosome-scaled cM assigned to the interval
• cM/Mb

Comparison of the Rodgers-Melnick and Ogut maps can be seen in RMv5_vs_ogut_cumulative_by_chromosome.png (scaled so cumulative value is the same).