BASEN: Genome-size results in detail

This page presents the analyses supporting Results section 1 of the main report (BASEN improves agreement with genome-copy composition as genome-size variation increases): L1 distance for every run against every reference composition, per-species estimates, the association between fold error and genome length, the yeast fold error per run, and a summary, for the three mock communities Microbial (D6300), Gut (D6331) and Oral (D6332). Each standard was classified against a Kraken2 database containing its corresponding catalogue genomes.

1. L1 distance per run, against every reference composition

L1 = Σ |estimated share − reference share| over the catalogue species
with both compositions as fractions summing to 1. It ranges from 0 (identical) to 2 (no species in common); half this distance represents the minimum fraction of total abundance that must be redistributed to match the reference (e.g., L1 0.336 = 17% abundance redistribution).

L1 distance to the manufacturer's composition (lower values indicate greater agreement) for each run and the six runs pooled, against every composition in the manufacturer's Table 1. Bold: lowest value in each column within each composition. Reads: Q ≥ 20 reads entering Kraken2.

• % cells differs from % genome copies mainly for the yeasts (two copies per cell; the bacteria shift by at most 0.3 percentage points after renormalization), resulting in the same method ranking as % genome copies.
• 16S copies: copies of the small-subunit ribosomal RNA gene (16S in bacteria and archaea, 18S in fungi such as the yeasts), whose copy number varies among species and is the target of amplicon sequencing. None of the four methods estimates gene copies; these comparisons are included to cover all manufacturer-reported reference measures. "Bacteria only" excludes the yeasts, which carry 18S rather than 16S; the profile is then renormalized over the bacteria.
• Microbial and Gut runs contained fewer reads (105–11,351 and 365–7,688, respectively) and showed greater variation in run-level L1 than Oral.

Microbial (D6300; six Flongle runs, reference set v3)

	Method	Run 1 327 reads	Run 2 247 reads	Run 3 1,223 reads	Run 4 1,725 reads	Run 5 105 reads	Run 6 11,351 reads	Pooled 14,978 reads	Mean ± SD runs 1–6
% gDNA
	Kraken2	0.256	0.405	0.608	0.211	0.386	0.159	0.173	0.337 ± 0.164
	Bracken	0.256	0.467	0.611	0.211	1.040	0.159	0.174	0.457 ± 0.333
	BASEN-nf	0.523	0.599	0.697	0.401	0.486	0.365	0.372	0.512 ± 0.124
	BASEN	0.516	0.601	0.709	0.401	0.482	0.361	0.369	0.512 ± 0.129
% 16S copies (bacteria only)
	Kraken2	0.368	0.471	0.567	0.324	0.521	0.421	0.393	0.445 ± 0.092
	Bracken	0.368	0.522	0.571	0.324	1.142	0.422	0.393	0.558 ± 0.301
	BASEN-nf	0.537	0.636	0.637	0.338	0.700	0.411	0.403	0.543 ± 0.143
	BASEN	0.541	0.638	0.650	0.337	0.705	0.410	0.403	0.547 ± 0.146
% 16S + 18S copies
	Kraken2	0.404	0.547	0.582	0.367	0.580	0.457	0.429	0.489 ± 0.093
	Bracken	0.404	0.632	0.585	0.367	1.252	0.458	0.430	0.616 ± 0.328
	BASEN-nf	0.672	0.737	0.738	0.452	0.788	0.535	0.528	0.653 ± 0.132
	BASEN	0.665	0.736	0.751	0.452	0.792	0.533	0.526	0.655 ± 0.135
% genome copies
	Kraken2	0.335	0.443	0.504	0.378	0.443	0.406	0.376	0.418 ± 0.059
	Bracken	0.335	0.496	0.507	0.378	0.978	0.406	0.376	0.517 ± 0.236
	BASEN-nf	0.506	0.601	0.541	0.397	0.633	0.337	0.336	0.503 ± 0.116
	BASEN	0.510	0.601	0.554	0.397	0.637	0.338	0.337	0.506 ± 0.117
% cells
	Kraken2	0.339	0.450	0.511	0.386	0.447	0.413	0.384	0.424 ± 0.059
	Bracken	0.339	0.495	0.515	0.386	0.971	0.413	0.384	0.520 ± 0.231
	BASEN-nf	0.508	0.605	0.545	0.395	0.631	0.336	0.335	0.503 ± 0.117
	BASEN	0.512	0.605	0.557	0.396	0.635	0.336	0.335	0.507 ± 0.118

• Limitation (small runs): BASEN's L1 to % genome copies was below Kraken2's only in run 6 (0.338 vs 0.406), which accounts for 76% of the pooled reads (11,351 of 14,978), so the pooled result largely reflects run 6 (pooled L1 0.337 vs 0.376). Run-level L1 is dominated by the bacteria, whose estimated abundances varied across runs for all methods (e.g., L. fermentum, expected 21.6% of genome copies, at 2.3–25.4% across runs and methods).

Gut (D6331; six Flongle runs)

	Method	Run 1 1,057 reads	Run 2 375 reads	Run 3 365 reads	Run 4 1,022 reads	Run 5 7,688 reads	Run 6 2,860 reads	Pooled 13,367 reads	Mean ± SD runs 1–6
% gDNA
	Kraken2	0.422	0.420	0.460	0.707	0.661	0.670	0.609	0.557 ± 0.136
	Bracken	0.436	0.429	0.510	0.743	0.661	0.681	0.609	0.577 ± 0.135
	BASEN-nf	0.306	0.334	0.384	0.703	0.718	0.786	0.686	0.538 ± 0.219
	BASEN	0.307	0.327	0.384	0.708	0.721	0.789	0.691	0.539 ± 0.222
% 16S copies (bacteria only)
	Kraken2	0.427	0.428	0.492	0.667	0.634	0.631	0.587	0.546 ± 0.110
	Bracken	0.439	0.439	0.515	0.687	0.634	0.635	0.588	0.558 ± 0.108
	BASEN-nf	0.179	0.298	0.430	0.644	0.660	0.726	0.630	0.489 ± 0.221
	BASEN	0.183	0.319	0.430	0.649	0.662	0.728	0.634	0.495 ± 0.218
% 16S + 18S copies
	Kraken2	0.422	0.443	0.475	0.712	0.684	0.682	0.627	0.570 ± 0.136
	Bracken	0.432	0.452	0.541	0.747	0.684	0.693	0.627	0.592 ± 0.135
	BASEN-nf	0.278	0.380	0.503	0.732	0.745	0.811	0.713	0.575 ± 0.219
	BASEN	0.283	0.416	0.504	0.738	0.747	0.814	0.717	0.584 ± 0.214
% genome copies
	Kraken2	0.518	0.532	0.575	0.815	0.781	0.777	0.730	0.666 ± 0.138
	Bracken	0.531	0.545	0.621	0.842	0.781	0.790	0.731	0.685 ± 0.136
	BASEN-nf	0.226	0.266	0.500	0.673	0.687	0.760	0.656	0.519 ± 0.228
	BASEN	0.229	0.299	0.503	0.678	0.690	0.763	0.660	0.527 ± 0.222
% cells
	Kraken2	0.523	0.536	0.577	0.820	0.785	0.780	0.734	0.670 ± 0.139
	Bracken	0.536	0.548	0.621	0.840	0.785	0.788	0.735	0.686 ± 0.134
	BASEN-nf	0.231	0.270	0.503	0.671	0.685	0.758	0.654	0.519 ± 0.225
	BASEN	0.233	0.301	0.507	0.676	0.687	0.760	0.658	0.527 ± 0.219

• Per run: BASEN's L1 to % genome copies was below Kraken2's in every run (0.229–0.763 vs 0.518–0.815), including the two lowest-depth runs (365 and 375 reads), indicating that low read counts alone do not explain the Microbial pattern.

Oral (D6332; six MinION runs on one flow cell)

	Method	Run 1 165,929 reads	Run 2 111,559 reads	Run 3 85,055 reads	Run 4 75,026 reads	Run 5 30,780 reads	Run 6 36,485 reads	Pooled 504,834 reads	Mean ± SD runs 1–6
% gDNA
	Kraken2	0.291	0.324	0.331	0.366	0.315	0.293	0.292	0.320 ± 0.028
	Bracken	0.291	0.324	0.331	0.367	0.315	0.293	0.292	0.320 ± 0.028
	BASEN-nf	0.328	0.322	0.320	0.315	0.332	0.322	0.322	0.323 ± 0.006
	BASEN	0.330	0.323	0.321	0.316	0.334	0.323	0.323	0.324 ± 0.007
% 16S copies
	Kraken2	0.325	0.353	0.360	0.402	0.337	0.318	0.319	0.349 ± 0.031
	Bracken	0.324	0.353	0.361	0.403	0.337	0.317	0.319	0.349 ± 0.031
	BASEN-nf	0.333	0.332	0.346	0.341	0.340	0.321	0.335	0.335 ± 0.009
	BASEN	0.334	0.333	0.346	0.341	0.341	0.322	0.336	0.336 ± 0.009
% genome copies
	Kraken2	0.323	0.359	0.366	0.408	0.341	0.316	0.320	0.352 ± 0.034
	Bracken	0.322	0.359	0.367	0.409	0.341	0.315	0.319	0.352 ± 0.034
	BASEN-nf	0.354	0.348	0.346	0.341	0.358	0.348	0.348	0.349 ± 0.006
	BASEN	0.356	0.349	0.347	0.342	0.360	0.349	0.349	0.350 ± 0.007
% cells
	Kraken2	0.323	0.359	0.366	0.408	0.341	0.316	0.320	0.352 ± 0.034
	Bracken	0.322	0.359	0.367	0.409	0.341	0.315	0.319	0.352 ± 0.034
	BASEN-nf	0.354	0.348	0.346	0.341	0.358	0.348	0.348	0.349 ± 0.006
	BASEN	0.356	0.349	0.347	0.342	0.360	0.349	0.349	0.350 ± 0.007

• Per run: genome sizes span only 1.4-fold, so the methods differ little: BASEN's L1 to % genome copies was below Kraken2's in runs 2, 3 and 4 and above it in runs 1, 5 and 6, but remained within 0.342–0.360 in every run, against 0.316–0.408 for Kraken2 (SD 0.007 vs 0.034).

2. Per-species estimates (six runs pooled)

Abundance estimates (%) are presented alongside the manufacturer's % gDNA and % genome copies, renormalized over catalogue species and reported to one decimal place. Genome: the reference genome length supplied to BASEN, in Mb (Gut E. coli: mean of its five strains). Bold: the estimate closest to % genome copies in each row (smallest difference in percentage points, as L1 measures; all ties are shown in bold, none when all four are equal). 0 = no reads assigned; – = Kraken2 assigned reads but fewer than Bracken's 10-read cut-off; 0.0 = under 0.05%. Species are listed in descending order of genome-copy abundance.

Microbial (D6300)

Species	Genome (Mb)	% gDNA	% genome copies	Kraken2	Bracken	BASEN-nf	BASEN
L. fermentum	2.23	12.0	21.6	8.3	8.2	5.3	5.4
S. aureus	2.73	12.0	15.2	13.3	13.2	16.1	16.0
E. faecalis	2.90	12.0	14.6	16.5	16.6	23.6	23.6
L. monocytogenes	3.03	12.0	13.9	10.4	10.5	13.8	13.8
B. subtilis	4.22	12.0	10.3	11.6	11.6	11.6	11.7
S. enterica	4.82	12.0	8.7	10.8	10.8	9.2	9.2
E. coli	4.93	12.0	8.5	13.3	13.3	13.0	13.1
P. aeruginosa	6.85	12.0	6.1	11.0	11.0	6.8	6.9
S. cerevisiae (yeast)	18.77	2.0	0.6	3.6	3.6	0.4	0.4
C. neoformans (yeast)	26.31	2.0	0.4	1.1	1.1	0.1	0.1

Gut (D6331)

Species	Genome (Mb)	% gDNA	% genome copies	Kraken2	Bracken	BASEN-nf	BASEN
V. rogosae	2.16	14.0	19.9	11.4	11.4	22.8	22.9
F. prausnitzii	2.91	14.0	14.8	24.7	24.7	23.2	23.3
R. hominis	3.46	14.0	12.4	4.3	4.3	3.4	3.3
L. fermentum	1.91	6.0	9.7	2.4	2.4	2.9	2.9
B. adolescentis	2.09	6.0	8.8	1.2	1.2	1.3	1.3
E. coli	4.95	14.0	8.7	33.1	33.1	30.1	30.1
B. fragilis	5.17	14.0	8.3	8.6	8.6	6.0	6.0
F. nucleatum	2.45	6.0	7.5	6.0	6.1	3.3	3.1
P. corporis	2.95	6.0	6.3	3.3	3.3	4.0	4.0
A. muciniphila	2.85	1.5	1.6	1.3	1.3	1.7	1.7
C. difficile	4.21	1.5	1.1	2.0	2.0	1.0	0.9
S. cerevisiae (yeast)	12.84	1.4	0.3	0.9	0.9	0.2	0.2
C. albicans (yeast)	12.58	1.5	0.3	0.7	0.7	0.1	0.1
M. smithii	1.85	0.1	0.2	0.0	–	0.0	0.0
S. enterica	4.76	0.0	0.0	0.0	–	0.0	0.0
E. faecalis	2.85	0.0	0.0	0	0	0	0
C. perfringens	3.44	0.0	0.0	0	0	0	0

Oral (D6332)

Species	Genome (Mb)	% gDNA	% genome copies	Kraken2	Bracken	BASEN-nf	BASEN
S. parasanguinis	2.14	16.0	17.1	16.8	16.9	12.7	12.7
V. parvula	2.18	16.0	17.0	17.0	17.0	21.9	21.9
N. subflava	2.20	16.0	16.1	16.0	16.0	20.9	21.0
P. nigrescens	2.68	16.0	12.8	21.1	21.1	17.3	17.3
S. mitis	1.87	8.0	9.1	5.6	5.6	5.9	5.9
H. parainfluenzae	2.14	8.0	8.8	7.8	7.8	10.0	10.1
S. odontolytica	2.46	8.0	7.6	1.5	1.5	2.0	2.0
R. dentocariosa	2.46	8.0	7.3	3.0	3.0	3.8	3.8
S. mutans	2.02	1.0	1.2	0.4	0.4	0.5	0.5
F. nucleatum	2.18	1.0	1.0	7.4	7.4	2.7	2.6
S. salivarius	2.20	1.0	1.0	2.3	2.3	1.0	1.0
P. gingivalis	2.40	1.0	1.0	1.0	1.0	1.3	1.3

• Shared deviations from the Gut reference composition: E. coli accounted for 30–33% of each estimated profile (14% of the DNA, 8.7% of genome copies), while B. adolescentis, R. hominis and L. fermentum were at 1.2–4.3% (6–14% of the DNA). These L1 deviations were shared across all four methods, consistent with effects preceding abundance estimation, including extraction, sequencing or the common Kraken2 classifications. The two members at ≤ 0.001% (E. faecalis, C. perfringens) had no assigned reads, and Bracken also excluded M. smithii and S. enterica (below its 10-read cut-off).

3. Fold error against genome length

Spearman rho between each species' genome length and its log fold error against % genome copies (log fold error = ln(estimate / expected); six runs pooled). Log transformation makes overestimation and underestimation symmetric: 2× and 0.5× are equidistant from 1× (log +0.69 and −0.69), and 1× is 0. Positive rho indicates that overestimation increases with genome length; values near 0 indicate little monotonic association with genome length. Bold: closest to 0 in each row. N: species with a non-zero estimate in every method.

Standard	n	Kraken2	Bracken	BASEN-nf	BASEN
Microbial (genome sizes 12-fold)	10	0.93	0.93	-0.09	-0.04
Gut (genome sizes 7-fold)	13	0.79	0.79	0.18	0.18
Oral (genome sizes 1.4-fold)	12	0.17	0.17	0.19	0.19
All three	35	0.59	0.59	0.06	0.06

• Kraken2 overestimation increased with genome length (rho 0.93 in Microbial, 0.79 in Gut); BASEN substantially reduced this association (-0.04 and 0.18). In Oral, where genome sizes span only 1.4-fold, neither method showed a comparable association (0.17 and 0.19).

4. Yeast fold error per run

Fold error expresses the ratio of estimated to expected abundance. A method that reports a species at 2% when 1% is expected has a fold error of 2× (twice the expected abundance); one that reports 0.5% has 0.5× (half the expected abundance); 1× indicates exact agreement.

fold error = estimate / expected % genome copies
(1× = exact; above 1 = overestimated; 0 = no reads assigned; – = Kraken2 assigned reads but fewer than Bracken's 10-read cut-off). Bold: closest to 1× in each column (by fold, so 0.5× and 2× are considered equally distant); all ties are shown in bold.

Microbial: S. cerevisiae (expected 0.57% of genome copies)

Method	Run 1 327 reads	Run 2 247 reads	Run 3 1,223 reads	Run 4 1,725 reads	Run 5 105 reads	Run 6 11,351 reads	Pooled
Kraken2	8.54×	3.52×	8.40×	5.12×	3.94×	6.26×	6.28×
Bracken	8.54×	–	8.35×	5.11×	–	6.25×	6.27×
BASEN-nf	1.31×	1.02×	0.98×	0.68×	0.03×	0.69×	0.72×
BASEN	1.26×	1.02×	0.95×	0.67×	0.03×	0.68×	0.71×

Microbial: C. neoformans (expected 0.37% of genome copies)

Method	Run 1 327 reads	Run 2 247 reads	Run 3 1,223 reads	Run 4 1,725 reads	Run 5 105 reads	Run 6 11,351 reads	Pooled
Kraken2	0	1.36×	4.06×	4.03×	0	2.95×	3.06×
Bracken	0	–	4.03×	4.02×	0	2.94×	3.05×
BASEN-nf	0	0.02×	0.16×	0.55×	0	0.34×	0.35×
BASEN	0	0.02×	0.17×	0.55×	0	0.34×	0.35×

Gut: C. albicans (expected 0.31% of genome copies)

Method	Run 1 1,057 reads	Run 2 375 reads	Run 3 365 reads	Run 4 1,022 reads	Run 5 7,688 reads	Run 6 2,860 reads	Pooled
Kraken2	8.89×	13.83×	8.56×	1.61×	1.23×	1.04×	2.32×
Bracken	8.99×	14.03×	–	–	1.23×	–	2.32×
BASEN-nf	2.38×	2.08×	1.21×	0.20×	0.20×	0.39×	0.41×
BASEN	2.32×	2.08×	1.22×	0.19×	0.20×	0.39×	0.40×

Gut: S. cerevisiae (expected 0.32% of genome copies)

Method	Run 1 1,057 reads	Run 2 375 reads	Run 3 365 reads	Run 4 1,022 reads	Run 5 7,688 reads	Run 6 2,860 reads	Pooled
Kraken2	10.20×	8.93×	12.91×	1.88×	1.77×	0.78×	2.68×
Bracken	10.32×	9.06×	13.67×	–	1.77×	–	2.68×
BASEN-nf	2.96×	1.69×	3.03×	0.65×	0.48×	0.11×	0.60×
BASEN	2.86×	1.75×	2.82×	0.65×	0.46×	0.11×	0.58×

• In pooled analyses, read-count estimates for the four yeasts were 2.3–6.3× and BASEN estimates were 0.35–0.71× the expected genome-copy abundances. BASEN was closer for S. cerevisiae in both standards; for C. neoformans (3.06× vs 0.35×) and C. albicans (2.32× vs 0.40×) the two methods showed errors of similar fold magnitude in opposite directions.
• In Kraken2, BASEN-nf and BASEN, Gut runs 1–3 yielded several-fold higher yeast estimates than runs 4–6; the cause remains unresolved.

5. Summary

• Pooled L1: BASEN-nf and BASEN were closest to % genome copies in Microbial (0.336 and 0.337 vs 0.376 for Kraken2) and Gut (0.656 and 0.660 vs 0.730); Kraken2 and Bracken were closest to % gDNA. In Oral, where genome sizes span only 1.4-fold, the four methods were within 0.031 of each other. % cells yields the same method ranking as % genome copies; none of the methods estimates 16S copies.
• Per run: BASEN was closer to % genome copies than Kraken2 in every Gut run, but in Microbial only in run 6, which accounts for 76% of the pooled reads; the Microbial pooled result therefore largely reflects one run. In low-depth runs, bacterial abundance estimates varied across runs for every method.
• Per species: BASEN-nf or BASEN yielded the estimate closest to % genome copies for 8 of 10 species in Microbial, 10 of 13 in Gut, and 7 of 12 in Oral (six runs pooled).
• Genome length: Kraken2 overestimation increased with genome length (rho 0.93 in Microbial, 0.79 in Gut); BASEN substantially reduced this association (-0.04 and 0.18).
• Yeasts: in pooled analyses, read-count estimates for the four yeasts were 2.3–6.3× and BASEN estimates were 0.35–0.71× the expected genome-copy abundances. Yeast estimates varied across runs for every method: the method closest to 1× differed across runs, and Gut runs 1–3 and 4–6 differ several-fold in Kraken2, BASEN-nf and BASEN, so the overall yeast comparison is based on pooled samples.