Single virus per well infections 20260730_single_well

Analysis of the barcode sequencing of 20260730_single_well_H3_strains (2026-07-30), 20260730_single_well_H1_vax_old_strains (2026-07-30), whose wells were each infected with a single virus of the library rather than with the pool. The plots are interactive: mouse over a bar for details.

Experimental description

Each well of these plates was infected with a single virus of the library rather than the pool, with the neutralization standard spiked in. Together they measure how well each strain grows on its own, and whether virus from one well ends up in another: a well should hold only its own strain and the neutralization standard, so anything else in it is contamination, and where that material comes from says whether it spilled between neighboring wells or arrived some other way.

Strains assayed here

151 wells across 2 plates hold one strain each, all at a dilution factor of 1. The barcodes are counted against flu-seqneut-2026-barcode-to-strain-designed.csv, which holds 154 strains.

3 of those strains are not assayed here, so there is no data on them. Whether each was carried through to flu-seqneut-2026-barcode-to-strain-actual.csv, the library used for the titer measurements, is given alongside:

• A/Bangkok/P2323/2025_H3N2 — not carried through either
• A/England/1845724/2025_H3N2 — not carried through either
• A/Galicia/GA-CHUAC-449/2025_H1N1 — not carried through either

3 of the 151 strains that are assayed were not carried through to flu-seqneut-2026-barcode-to-strain-actual.csv. Their bars are outlined and their table rows written in this color throughout, so that what was measured here can be read against the decision to drop them:

• A/California/LACPHL-INF02113/2025_H3N2
• A/Netherlands/1739/2023_H1N1
• A/Netherlands/446/2026_H1N1

A further 4 strains were carried through with fewer barcodes than were counted here, which matters because a strain's reads are summed over its barcodes. These count as carried forward and are not marked:

• A/Hawaii/70/2019_H1N1 — 4 barcodes counted here, 3 carried through
• A/Peru/ANC-INS-062/2026_H3N2 — 2 barcodes counted here, 1 carried through
• A/SouthAustralia/2605715552/2026_H3N2 — 2 barcodes counted here, 1 carried through
• A/Valencia/VAHNSI_09_02564/2025_H1N1 — 2 barcodes counted here, 1 carried through

Reads per well and what each well held

One row per well, across every plate of this group, labeled by the strain the well was infected with. Hovering a bar outlines that strain in all three panels, and its tooltip names the plate and well the row is from.

The left panel is what became of every read sequenced for the well. Where the reads went:

• valid barcode: the barcode was read and is in the viral library or the neutralization standard set, so the read is counted below
• invalid barcode: the barcode was read but is in neither, so the read counts towards neither strain nor standard; these are the reads the unmatched barcodes section is about
• low quality barcode: the barcode was found but some of its nucleotides were sequenced too poorly to be trusted
• unparseable barcode: the sequence flanking the barcode was found but the barcode itself could not be read from the position it should be in
• invalid outer flank: the outer flanking sequence did not match. Where an outer barcode is used to separate samples that share a sequencing run, reads belonging to the other samples are what fails this way, so some of these reads are expected rather than a problem

The middle panel is what the well's virus was, as fractions of the reads that could be parsed as having a barcode, regardless of whether or not that barcode is supposed to be in the viral library, excluding neutralization standard barcodes. The four categories partition those reads and so sum to one:

• own strain: a barcode of the strain this well was infected with.
• other library strains: a barcode of a different strain of the library, which does not belong in this well at all.
• within 1 nt of strain's or neut standard barcode: matched no known barcode, but is that close to one belonging in this well, so it is a misread of something that belongs rather than contamination.
• over 1 nt from strain's or neut standard barcode: matched no known barcode and is further away than that, so it cannot be put down to either. This holds other strains' material misread as well as barcodes foreign to the library outright.

The right panel takes the other library strains of the middle panel and says where they came from, on the same fraction scale, so its bars are small by construction. Every read of another strain is attributed to the well that strain was grown in, and to how far that well is from the one the read turned up in; a step onto any of the eight surrounding wells counts as one, diagonals included, so wells one apart are the wells that touch. Material one well away is liquid carried between neighboring wells. Material from further off, from a well on another plate of this group, or from a strain grown nowhere here has to have arrived some other way, and distances above 2 wells are grouped together since there is nothing to tell them apart.

All three panels are ordered by how much of a well is not the own strain's material at all, largest first: the other library strains plus the further unmatched reads. Reads misread off a barcode that belongs in the well are deliberately left out of that sum. Read the left panel's counts alongside the fractions, as a well whose own strain barely grew shows a large fraction of foreign material from a very small amount of it.

Wells holding another strain

The wells where more than 0.005 of the parsed reads are barcodes of another library strain, with the 3 strains contributing most of them. Read the counts alongside the fraction: a well whose own virus failed to grow shows a large fraction of another strain from very little of it, and a shallow well reaches this threshold on very few reads, which is why parsed reads is here too.

Barcodes not matching any strain or neut standard

Reads that parsed as a barcode but match neither the viral library nor the neutralization standard. One within 1 nucleotide of a known barcode is sequencing error off it; one further away is a genuinely different barcode, and so material that is not in the library at all.

A well whose virus grew well but whose barcodes are not in the library being counted against would show up here and nowhere else, its reads all landing in this category while the well looks empty of virus everywhere else in the report.

The wells where more than 0.02 of the parsed reads matched neither set, one row each, with the 3 such barcodes the well holds most of. Each is given as its share of the well's non-neut-standard reads, followed by what it appears to be: the barcode of something known, misread, or a sequence far enough from everything known to be foreign to the library.

Estimated titer of each strain

An estimate of how well each strain grew on its own: its own reads against the neutralization standard spiked into its well, own strain reads / neutralization standard reads. Only those two are in the ratio, so a well holding material that does not belong does not get a higher titer for it; that is reported above instead.

Read it as an estimate rather than a measurement. It is relative, in arbitrary units rather than infectious ones, saying how much of a strain there was for a fixed amount of standard at a dilution factor of 1. And the infection may not have been in the range over which barcode counts track how much virus a well held, in which case the ratio understates or overstates the difference between two strains rather than scaling with it. Reads are also summed over all of a strain's barcodes, of which strains have differing numbers, so a strain carrying more barcodes gives more reads at the same titer; the barcode count is in each point's tooltip.

The standard's reads are this ratio's denominator, and so are what sets how precisely it is measured. A strain whose well holds fewer than 100 of them is named in a warning at the top of this report, its titer being imprecise however the point sits; this is the one place in the report where the amount of standard matters quantitatively. It only warns: the strain keeps its point, since a titer that is imprecise is still a measurement.