Additional file 1 of Long-read sequencing and genome assembly of natural history collection samples and challenging specimens

Additional file 1: Fig. S1: Levels of exogenous DNA contamination. For the samples sequenced with the PacBio ultra-low library protocol, we assembled the HiFi reads and screened the contigs for exogenous DNA contamination. The left plots show the total length of sequences not flaggedor flaggedas contamination by FCS. The right plots show the total length of sequences assigned to a certain contaminant taxa. Sequences tagged by the FCS status “exclude” and “trim” show clear non-endogenous taxonomic signals, while sequences tagged with “review” are borderline cases.For contig assembly of reads obtained from the Idiurus macrotis sample, most of the sequenced DNA is contamination from bacteria. Aligning the sequenced reads to the genome of a close relativerevealed an extremely low mapping rate of < 0.1%, confirming that essentially no endogenous DNA is in the sample.For contig assembly of reads obtained from the Anguis fragilis sample from 2021, the majority of the assembled sequence is putatively endogenous DNA; however bacterial contamination is also present. Aligning the sequenced reads to the genome of a close relativeshowed a moderate mapping rate of 36,27%, confirming that a part of the sequenced DNA is endogenous.For the final chromosome-level assembly of the maned sloth, no contamination was found. FCS identifies very low levels of plant contamination. Consistent with these results for the final assembly, screening the HiFi reads of the first SMRT cell with kraken2 [54] indicated very low levelsof non-mammalian DNA in the sample.The Elysia timida hifiasm-phased assembly of haplotype 1, generated from HiFi data obtained with both polymerases A/B and C, exhibits low levels of contamination mostly coming from bacteria. Fig. S2: Breakdown of taxa contributing to exogenous DNA contamination. Blobplots created with blobtoolkit [44] for assemblies created with the ultra-low input protocol. The central plot depicts assembly contigs as blobs, with the blob sizes corresponding to contig sizesand colors representing the assigned taxon as inferred from the best blast hits. The Y-axis shows the coverage of sequenced reads mapped back to the respective contig and the X-axis shows the average GC-content of each contig. Histograms on top and right hand side summarize average GC-content and coverage of assembly contigs, partitioned by the assigned taxa.The contig assembly of Idiurus macrotis reads shows that most of the assembled contigs are assigned to different bacteria, and coverage as well as GC content of assembled contigs varies widely, indicating that the assembly mostly consists of contaminant sequence.For the contig assembly of the Anguis fragilis sample from 2021, the majority of the assembly is assigned to either “Chordata”or “no-hit”, and “no-hit” contigs partly resemble “Chordata” contigs in terms of coverage and GC-content. On the other hand, the longest contigs in this assembly are assigned to “Pseudomonadota”, meaning that there is evident bacterial contamination present in the assembly. This sample was collected as a roadkill, and partial decomposition by bacteria provides an explanation for the observed contamination.In the final contig level assembly of Bradypus torquatus, only two small contigs are assigned to “Streptopytha”, highlighting the absence of systematic contamination.The majority of contigs of the Elysia timida hifiasm-phased assembly of haplotype 1 is assigned to “Mollusca”, but the blobplot clearly shows several other clusters consisting of large contigs assigned to “Pseudomonadota”and other bacterial taxa or “no-hit”. As the individual sequenced here was immediately killed and homogenized for sequencing without prior fixation, sample decomposition is unlikely. Rather, Elysia timida is too small to dissect before sequencing, therefore the observed contamination is likely caused by the gut microbiome and other commensal microorganisms. It should be noted that mollusca are underrepresented in the blast database, which likely increases false-positive hits and “no-hit” assignments [66]. Fig. S3: Amplification with polymerase C creates more even read coverage across exons and repeats.Violin plots depict the distribution density of normalized read coverage, calculated as the read coverage at a given genomic position divided by the average coverage of the respective library. Horizontal lines show 25% quartile boundaries, median and 75% quartile boundaries, respectively. Dotted lines represent a normalized coverage of 0.25. Only genomic regions annotated as exons or repeats were considered. Reads from polymerase C cover exonic and repeat regions more evenly, as the mode of the distribution is more pronounced around ~ 1.0 normalized coverage.Number of unique exons having ≤ 0.25 normalized coverage. The overall number of unique exons is 256,011, 246,006, and 239,251. There is an up to ten-fold excess of exons having ≤ 0.25 coverage for polymerase A/B compared to polymerase C, highlighting potential drop-outs that can lead to a fragmented assembly and missing annotations.Bradypus torquatus sequencing reads obtained with polymerase A/B or C, aligned against the Choloepus didactylus genomeor the Bradypus torquatus genome.Human sequencing data aligned against the HG002v1.1 maternal haplotype assembly, comparing reads obtained with polymerase A/B or C and reads obtained without amplification. Fig. S4: Low fraction of chimeric reads in amplified PacBio HiFi-read libraries.The Y-axis shows different types of alignments as classified by minimap2 [64] when mapping human reads to the HG002v1.1 assembly. The X-axis depicts the fraction of alignments from a library. Supplementary alignments of a read indicate that the read is a chimer consisting of two different genomic regions. Alignments of non-amplified ultra-low input reads serve as a baseline and show a similarly low fraction of supplementary alignmentsas reads obtained with polymerase A/Band polymerase C. In stark contrast, the majority of MDA reads have supplementary alignments and are likely chimeric.Histograms of raw read and alignment lengths of different PacBio libraries, following the color scheme of A. The filled histogram areas indicate raw read length. The red and blue lines show primary and supplementary alignment lengths excluding soft- and hard-clipped bases. Whereas primary alignment lengths closely follow raw read length distributions for ultra-low and non-amplified reads, indicating full length mapping, for MDA reads, the mode of the primary alignment distribution is located at around ~ 5000 nt compared to ~ 11,000 nt length for raw reads, clearly showing that primary alignments tend to be severely truncated. In accordance with, the amount and length of supplementary alignments is very low in non-amplified and ultra-low reads, but exceptionally high in MDA reads. Fig. S5: Both polymerase A/B and C exhibit PCR bias for Elysia timida. IGV screenshots show alignments of HiFi reads produced with the ultra-low input protocol using either polymerase A/B or polymerase C. Reads were aligned to an assembly of Elysia timida generated with all reads from all three polymerases.Two examples of genomic regions where HiFi read coverage is very low or drops to zero for polymerase A/B while polymerase C reads cover the region. This exemplifies regions difficult to sequence with polymerase A/B.Two examples of loci where read coverage drops to zero for polymerase C but not polymerase A/B. While inreads from polymerase A/B still cover the locus, it should be noted that there are fewer reads compared to the flanking regions. This indicates that it is also difficult for polymerase A/B to amplify these genomic regions. Fig. S6: Polymerase A/B and C biases for Podura aquatica and Scintilla philippinensis. IGV screenshots show alignments of HiFi reads produced with the ultra-low input protocol using either polymerase A/B or polymerase C. Reads were aligned to assemblies of Scintilla philippinensisand Podura aquaticagenerated with all reads from polymerases A/B and C.Two examples of genomic regions where HiFi read coverage is very low or drops to zero for polymerase A/B, while polymerase C reads cover these regions. This exemplifies regions difficult to sequence with polymerase A/B.Two examples of loci where read coverage drops to zero for polymerase C but not polymerase A/B. Of note, in panel D, polymerase A/B coverage is also comparatively low, and in panel B, there is a polymerase A/B coverage dropout upstream of the polymerase C dropout. Together, this illustrates that for certain genomic regions one or both polymerases have difficulty amplifying DNA. Note 1: PacBio ultra-low library preparation based on PCR amplification with KOD Xtreme™ Hot Start DNA Polymerase (Merck). Note 2: Commands used for scaffolding the Bradypus torquatus assembly.