How to use this explorer

open the explorer what the v2 analysis changed

This page explains what the explorer measures, what every control does, and how to read every number on the screen. You do not need to know the analysis to use it. Read the first two sections before you start, then use the rest as a reference.

What this tool answers

The explorer reports how often resistance to antibiotics and resistance to metals or biocides are carried together in the same bacterial genome. It covers 59,626 environmental genomes. It lets you ask that question along four axes at once, so you can hold three of them fixed and read the fourth.

A typical question it answers: in wastewater Klebsiella pneumoniae, how often does a mercury resistance gene sit on the same plasmid as a carbapenem resistance gene.

The two mechanisms it keeps apart

Published co-occurrence figures often mix two different things. This tool reports them separately, because they are measured on different units and they mean different things. The Component switch in the rail chooses which one you are looking at.

Cross-resistance

One single protein carries both an antibiotic annotation and a metal or biocide annotation. One gene product can act on both kinds of agent, so no gene transfer is needed for the two resistances to travel together. A multidrug efflux pump that exports both a quaternary ammonium biocide and an antibiotic is the common case. 42,469 of the 59,626 eligible genomes carry such a protein, which is 71.23 percent. Cross-resistance is counted in genomes.

Co-resistance

Two different genes sit on the same mobile element: one antibiotic resistance gene and one separate metal or biocide resistance determinant. The element can move both genes together into a new host. 6,377 genomes carry such an element when plasmids and ICEfinder elements are both counted, which is 10.69 percent. Co-resistance is counted in elements as well as genomes, because one genome can carry several such elements. There are 7,016 of those elements.

The two figures are not comparable. Cross-resistance is high because the proteins that satisfy it are broad substrate efflux pumps that most of these organisms already carry. Co-resistance is the measure that reflects two determinants brought together on something that moves.

The denominator, and what a percent means here

Every percentage on the screen is printed next to the count it was computed from, so you can always see what the denominator is. Check it before you compare two numbers.

An eligible genome carries at least one antibiotic resistance gene and at least one metal or biocide determinant, anywhere in the genome, on any replicon. There are 59,626 of them. Everything in the tool is computed on eligible genomes, so a genome with no metal or biocide determinant never sits in a denominator as a negative.

The denominator changes with the component you chose. In the cross component, shares are taken over cross-resistant genomes. In the co component, shares are taken over co-resistant elements. The panel titles say which, and the matrix legend restates it.

Present and associated are different counts, and the tool shows both. Present means the determinant occurs somewhere in the genome. Associated means it sits in a protein that also holds an antibiotic annotation, in the cross component, or on a mobile element that also holds an antibiotic gene, in the co component. Globally 110 compounds and 35 drug classes are present, while 70 compounds and 24 drug classes are associated in the cross component.

The four views

ButtonWhat it showsUse it to
OverviewSix panels covering the whole current selection at once Get your bearings and see where the signal sits
SelectionOne species and one compartment in detail, with ranked lists Answer a question about one organism in one place
MatrixEvery compound against every drug class as a heatmap Find which specific pairs occur, and which never do
GraphA rotatable three dimensional view with four rings of nodes See how one node connects across all four axes

The views share one selection. Changing a filter changes every view at the same time.

The filter rail, control by control

Component: Cross or Co

Chooses the mechanism, as described above. This is the most important control on the page, because it changes what every number means.

Co counted on: Plasmids, or Plasmids and ICEs

Visible only when the Co component is selected. Plasmids counts pairs on MOB-Suite plasmid clusters, which gives 5,380 genomes and 5,887 elements. Plasmids and ICEs adds the integrative and conjugative elements that ICEfinder reports, which gives 6,377 genomes and 7,016 elements. ICEfinder assessed 59,622 of the 59,626 genomes, so the wider setting costs almost nothing in coverage.

Compound groups: Metals and biocides, or All groups

Metals and biocides is the default and shows 63 of the 110 compounds. All groups adds the other 47, which are dyes, detergents, solvents, bile salts, fatty acids and lipophilic probes. Those 47 are chemicals an environmental genome is not usually exposed to as a contaminant, so the default leaves them out. Switch them on to compare with work that counted them.

Disinfectant drug class: Hide or Show

Hidden by default. The CARD drug class named "disinfecting agents and antiseptics" is carried by qacE and qacEdelta1, and those same genes are the biocide determinant, so pairing that class with a biocide compound is circular. Showing it restores the class to every panel. No genome's eligibility depends on it, because no genome has a qac gene as its only antibiotic determinant.

Compartment

Restricts every panel to one compartment, or to Global for all of them together. The five compartments hold 22,381 water, 21,428 unclassified, 9,056 wastewater, 4,342 soil and 2,419 sediment genomes.

Species

Restricts every panel to one species, or to All for every species together. Type in the box above the list to filter the list itself. Species with fewer than 20 eligible genomes are pooled into one bucket named "Other species (<20 genomes)", so that a share is never drawn on a handful of genomes. The species by compartment grid is the exception and lists all 154 species separately.

Metal or biocide, and Drug class

These two pin the view to one compound, one drug class, or one of each. Pinning both answers a single question: how often does this exact pair occur. The graph, matrix, compartment profile and CSV download then show only the pinned selection. The population panels keep the whole stratum, so you can still see the context. The hint under the control says how many of the two are pinned.

Display filter: hide cells below

Set to 5 by default. A compound or drug class earns a row, a column or a node only if at least one of its cells reaches this count. Raising it clears away single observations. It never changes the counts themselves: once a row and a column are shown, every cell in them shows its own true value, including values below the threshold.

Reset view, and Download view

Reset returns every control to its default. Download writes the current view to a CSV file, described below.

The panel at the foot of the rail

Prints the four counts for the current selection: eligible genomes, cross-resistant genomes, co-resistant genomes and co-resistant elements. When Plasmids and ICEs is selected it also prints how many of the genomes ICEfinder assessed.

The number strip at the top

Seven figures for the current selection. The first is the eligible genome count. The next four split those genomes into four classes that always sum to 100 percent: cross only, both, co only, and neither. The last two say how many drug classes and compounds are associated, out of those present.

Globally, with co counted on plasmids, the four classes are 37,110 cross only, 5,359 both, 21 co only and 17,136 neither. The 21 in co only matter: those genomes carry two determinants on one element without any single protein carrying both.

If the selection holds fewer than 20 eligible genomes, the word "small sample" appears beside the count. Treat the shares in that selection as unstable.

Reading the Overview panels

Architecture of eligible genomes

One stacked bar showing the same four classes as the number strip, with the count for each printed in the key below it. The line under the key gives cross in total and co in total, which are the two overlapping groups rather than the four exclusive ones.

By compartment

One bar per compartment, giving the share of that compartment with the selected component. The title names the denominator. When nothing is pinned it reads the share of eligible genomes with cross-resistance or co-resistance. When a compound or a drug class is pinned it reads the share of cross-resistant genomes, or of co-resistant elements, that carry the pinned item.

When the Co component and Plasmids are both selected, a second set of bars appears below. It counts co-resistant plasmids by mobility class, which tells you whether the elements carrying the pairs can transfer themselves.

Species by compartment

A grid with one row per species and one column per compartment, plus a column for all compartments together. Each cell is the percent of that species' eligible genomes in that compartment that carry the selected component. This grid ignores the species filter, the compartment filter and the display threshold, so it always shows the full picture.

Read the cells as follows. A blank cell means no genome in that stratum carries the component, which is a measured zero. NA means the stratum holds no eligible genome at all, so nothing was measured; hover the cell and the tooltip says which of the two it is. <1% means a value above zero and below one percent. A value printed in grey rests on fewer than 20 eligible genomes. Click a species name to run a web search for it. Click a value to open that species and compartment in the Selection view.

Top partners

The ten compounds and the ten drug classes that most often carry the selected component, with the share of the current denominator beside each. The title names that denominator.

The two presence grids

The last two panels give the percent of eligible genomes in which each compound's determinant, and each drug class, is present. Present means anywhere in the genome, with no co-location required, so these are the background rates the association panels sit against. They hold the same measurement as Figure 1 and Extended Data Fig. 1 of the manuscript, on the same denominator of 59,626, so the values match those figures.

Reading the Selection view

This view answers a question about one species in one compartment. The heading names the species, the compartment and the component. Four counts follow: how many drug classes are associated out of those present, how many compounds are associated out of those present, the eligible genome count, and the count of cross-resistant genomes or co-resistant elements.

The two lists below rank every associated drug class and every associated compound, with the count and the share. These lists show every associated determinant, including those below the display threshold, so nothing is hidden from them.

Reading the Matrix view

The matrix is the joint measurement, and it is the panel that answers which specific pairs occur. Rows are compounds and columns are drug classes. A cell is the number of genomes, in the cross component, or elements, in the co component, in which that compound and that drug class are carried together on one protein or on one element.

A row or a column appears only if one of its cells reaches the display threshold. Within the shown rows and columns, every cell is drawn at its own value. Hover any cell for the exact count, the denominator and the percent.

Grey means never observed: that pair was never seen on one protein, in the cross component, or on one element, in the co component. The colour ramp begins above zero, so the palest shade means present and rare. Under the default cross view, 494 of the 700 cells shown are grey, and that emptiness is the main signal of the panel.

Click a row or column header to run a web search for that compound or drug class.

Reading the Graph view

Four rings of nodes: species, compounds, drug classes and compartments. An edge between a compound and a drug class is the joint measurement, the same number the matrix holds. When a species or a compartment is selected or hovered, that node draws its own edges to every compound and drug class it carries.

Edges that run from a compound or a drug class out to a species or a compartment are marginals. They say the node occurs there. They do not say the pair occurs there. For a pair in a place, read the matrix with that place selected.

Drag to rotate, scroll to zoom, and click a node to follow it. Clicking a species or a compartment sets the filter. Clicking a compound or a drug class highlights its partners. Shift and click runs a web search on the node name.

Five questions, answered step by step

How many antibiotic classes does Escherichia coli carry in wastewater

  1. Type Escherichia coli in the search box at the top and pick it.
  2. Choose Wastewater in the Compartment list in the rail.
  3. Read the two counts at the top of the Selection view, with the ranked lists below them.

Which metal travels with carbapenem resistance on plasmids

  1. Set Component to Co and leave Co counted on at Plasmids.
  2. Open the Matrix view.
  3. Find the carbapenem column and read down it. The darkest cells are the metals that most often share a plasmid with a carbapenem resistance gene.
  4. Hover a cell for the element count, the denominator and the percent.

Is a pair concentrated in one compartment

  1. Pin the compound in the Metal or biocide list and the class in the Drug class list.
  2. Leave Compartment at Global.
  3. Read the By compartment panel in the Overview. Its title now names the pinned pair, and the bars give the share of each compartment that carries it.

Does this species carry the pair anywhere at all

  1. Set the Component you want.
  2. Find the species row in the Species by compartment grid.
  3. A blank cell is a measured zero, and NA means that stratum held no eligible genome. Hover the cell to see which.

Is a result resting on enough genomes to trust

  1. Read the eligible genome count in the number strip for the current selection.
  2. If the words "small sample" appear, the selection holds fewer than 20 genomes.
  3. In the species grid, a value printed in grey rests on fewer than 20 eligible genomes.
  4. Raise the display threshold to drop single observations from the matrix and the graph.

Taking the data out

Download view writes the cells of the current view to a CSV file. The file is named for the component, the species and the compartment. It holds one row per cell, with these columns:

ColumnMeaning
layercross or co, the component you had selected
co_defpl for plasmids or pi for plasmids and ICEs, empty in the cross component
species, compartmentThe selection the cell belongs to
compound, drug_classThe pair, under their stored names
nGenomes in the cross component, elements in the co component
denominator, pctThe denominator in force and the percent of it
unitgenomes or elements, so the row reads correctly on its own

The file carries the stored names, without the display shortening described below. The wider table behind the tool is results/tables/strata_summary.csv, which holds one row per species and compartment with every denominator.

Names you will see

Drug class names drop the trailing word antibiotic on screen, so "aminoglycoside antibiotic" reads as "aminoglycoside". Compounds show their category after a dot, so "Mercury (Hg)" reads as "Mercury (Hg) . metal". The stored names are untouched and the CSV uses them as they are.

Species names come from Check_M_Passed_Final_List_with_Species.xlsx. Where that workbook identified a genome only to a species group or to a genus, the name reads as the genus followed by sp. There are 79 such genomes and all of them read Bacillus sp. Every genome in the tool is named, so you will not find an unassigned bucket on any axis.

What these numbers support

These panels record carriage and co-location. No metal or biocide concentration is measured anywhere in this study. Nothing here describes exposure, and the design does not support causal or selection language.

Counts are genomes or elements. Plasmids inside one MOB-Suite cluster are near identical, so the counts hold repeated observations of the same element. No significance test is reported here or in the study, and none should be read into a difference between two cells.

Metal and biocide determinants come from BacMet2, experimentally confirmed entries only. Detection is uneven across genera. BacMet2 returns no hit in any screened Legionella genome, 0 of 1,405, or any screened Clostridium genome, 0 of 203. It hits 5.8 percent of Shewanella, 20.5 percent of Aeromonas and 69.1 percent of Bacillus, and 92 percent or more of every other genus with 20 or more screened genomes. In those first taxa, absence means absence of an experimentally confirmed determinant in the reference database. Comparisons between genera carry this bias. Comparisons within a genus and across compartments are free of it.

The co-resistance figure here counts pairs on plasmids and on ICEfinder elements. The manuscript reports 6,378 genomes and 10.70 percent because it also counts integrons and insertion sequences as elements. One genome separates the two figures, and it carries its only co-resistant pair on an integron.

The unclassified compartment is not a place. It holds genomes whose recorded isolation source did not map to water, wastewater, soil or sediment. Read it as missing information about the source.

Where the numbers come from

Antibiotic resistance genes come from CARD and metal and biocide determinants from BacMet2, both by protein level annotation. Plasmids come from MOB-Suite and integrative elements from ICEfinder. Genome metadata, including species and compartment, comes from Check_M_Passed_Final_List_with_Species.xlsx alone.

Every figure in the tool is built from the same analysis tables that produced the manuscript figures, so the shared numbers agree. The header of the explorer prints the release label and the build date. README.md in the project folder holds the full definitions, the species census and the rebuild instructions.

Caution. Association only. Read every percentage next to the denominator printed beside it, and keep the two components apart when you quote a figure.