Lake Malawi cichlid pangenome graph reveals extensive structural variation driven by transposable elements.

BackgroundMost DNA methylation studies have used a single reference genome with little attention paid to the bias introduced due to the reference chosen. Reference genome artifacts and genetic variation, including single nucleotide polymorphisms (SNPs) and structural variants (SVs), can lead to differences in methylation sites (CpGs) between individuals of the same species. We analyzed whole-genome bisulfite sequencing data from the fetal liver of Angus (Bos taurus taurus), Brahman (Bos taurus indicus), and reciprocally crossed samples. Using reference genomes for each breed from the Bovine Pangenome Consortium, we investigated the influence of reference genome choice on the breed and parent-of-origin effects in methylome analyses.ResultsOur findings revealed that ∼75% of CpG sites were shared between Angus and Brahman, ∼5% were breed specific, and ∼20% were unresolved. We demonstrated up to ∼2% quantification bias in global methylation when an incorrect reference genome was used. Furthermore, we found that SNPs impacted CpGs 13 times more than other autosomal sites (P < 5 times {10}^{ - 324}) and SVs contained 1.18 times (P < 5 times {10}^{ - 324}) more CpGs than non-SVs. We found a poor overlap between differentially methylated regions (DMRs) and differentially expressed genes (DEGs) and suggest that DMRs may be impacting enhancers that target these DEGs. DMRs overlapped with imprinted genes, of which 1, DGAT1, which is important for fat metabolism and weight gain, was found in the breed-specific and sire-of-origin comparisons.ConclusionsThis work demonstrates the need to consider reference genome effects to explore genetic and epigenetic differences accurately and identify DMRs involved in controlling certain genes.

Whole genome sequencing has revolutionized infectious disease surveillance for tracking and monitoring the spread and evolution of pathogens. However, using a linear reference genome for genomic analyses may introduce biases, especially when studies are conducted on highly variable bacterial genomes of the same species. Pangenome graphs provide an efficient model for representing and analyzing multiple genomes and their variants as a graph structure that includes all types of variations. In this study, we present a practical bioinformatics pipeline that employs the PanGenome Graph Builder and the Variation Graph toolkit to build pangenomes from assembled genomes, align whole genome sequencing data and call variants against a graph reference. The pangenome graph enables the identification of structural variants, rearrangements, and small variants (e.g., single nucleotide polymorphisms and insertions/deletions) simultaneously. We demonstrate that using a pangenome graph, instead of a single linear reference genome, improves mapping rates and variant calling for both simulated and real datasets of the pathogen Neisseria meningitidis. Overall, pangenome graphs offer a promising approach for comparative genomics and comprehensive genetic variation analysis in infectious disease. Moreover, this innovative pipeline, leveraging pangenome graphs, can bridge variant analysis, genome assembly, population genetics, and evolutionary biology, expanding the reach of genomic understanding and applications.

In this era of high-throughput genomics, the number of available genome sequences and genomic data sets for many organisms is growing rapidly. Many species are now represented by multiple genomes, rather than a single reference genome. Traditional comparative genomics approaches are bound by the physical constraints of a single genome that are unable to represent all genomic diversity of a species or population. As a result, there is an ongoing shift towards pangenomic approaches, which promise to precisely address the sequence and variation described in large collections of related genomes. However, the lack of practical applications of pangenomes prevents us from understanding the full genetic diversity variation within a species or population. PanTools is a recently developed toolkit for pangenome construction, graph annotation, and homology detection. In this thesis, I develop and employ novel applications utilizing PanTools' annotated pangenome graph representation, mainly focusing on the analysis of plant and bacterial pathogen genomes. In Chapter 1 I introduce the field of comparative genomics and explain the essential principles for comparing genetic blueprints of different organisms. I describe the limitations of traditional methods, highlighting how reference genome bias and computational constraints are driving a transition to pangenomic approaches. A brief historical overview of pangenome studies elucidates the significance of this transition and introduces novel concepts used in pangenomics. The most prevalent pangenome representations are further detailed, discussing their advantages and shortcomings. Finally, I delineate several critical challenges that must be overcome to advance the recently emerged field of pangenomics. Chapter 2 lays the groundwork for comparative genomics in PanTools by introducing a set of functionalities designed to handle large collections of bacterial genomes. Quality assurance to validate the genome quality and annotations before analyzing the data. Characterizing a pangenome's gene content (core, accessory, unique) and pangenome openness estimation. Furthermore, implementation of phylogenomic methods to reconstruct evolutionary relationships across bacterial species. The hierarchical graph representation was extended to enable incorporation of phenotype/metadata and functional annotations. The functionalities are demonstrated in a comparative study of nearly 200 Pectobacterium genomes. We describe the pangenome's large accessory genome, whose genetic repertoire far surpasses that of a single bacterial genome. The gene-based pangenome representation is further utilized as a framework for identifying genetic differences between blackleg causing and non-causing strains. In Chapter 3, the methodology introduced in the previous chapter is updated to enable large-scale comparative genomics in pangenomes throughout all kingdoms of life. To obtain an accurate representation of gene families across different genome collections with varying sizes, complexities, and evolutionary distances, we propose a method that adjusts strictness of the homology grouping based on validated single-copy orthologs. We demonstrate generic applicability and scalability of the software by emphasizing functional annotation, gene-level and phylogenomic analyses on seven use cases from different taxonomic kingdoms: ranging from viruses to human and plant genomes. Recent breakthroughs in sequencing and assembly technologies have delivered the first high quality haplotype-resolved (phased) genome assemblies. Chapter 4 focuses on harnessing the variation in these novel assemblies, which was previously hidden within haploid references. PanTools' methodology is updated, and novel functionalities are introduced specifically to enable the analysis of intragenomic variation, complementing existing intergenomic analyses. A relational synteny layer is incorporated into the hierarchical pangenome graph, linking collinear gene loci across all sequences. The functionalities are employed to pangenomes of diploid apples and tetraploid potatoes. We demonstrate high heterozygosity among homeologous chromosomes, characterized by gene absence/presence variation. Additionally, the analyses shed light on the evolutionary history of apple and potato species by revealing structural chromosomal rearrangements that occurred after whole-genome duplication events. Chapter 5 presents various enhancements to the PanTools toolkit required for integrating pangenomics into plant breeding applications. We address a major bottleneck in the pangenome construction algorithm to increase scalability for a higher number of genomes. Improved construction runtime is demonstrated on various datasets, with the largest being a Solanum (genus-level) pangenome, including a combined total of 29 (haplotype-resolved) tomato and potato genomes. We further introduce functionalities for detecting gene presence/absence variation in WGS samples and integrating them into the pangenome graph, showcasing gene-bases pangenomic analysis of 474 Lactuca resequencing lines. Finally, we introduce a visual analysis browser for exploring sequence variants in the context of 20 tomato references with metadata and 444 resequenced accessions. Chapter 6 serves as the conclusion of this thesis. I explain how the field of pangenomics has evolved during the timeframe of this thesis, detailing its current representations and applications, and how my work contributes to the field. Recognizing pangenomics as an emerging field that requires continuous development, I discuss the key challenges it faces. Finally, I conclude by exploring promising future directions for pangenomics research, with a particular focus on applications in bacteria and plants.

Analysis of sequence diversity in the human genome is fundamental for genetic studies. Structural variants (SVs) are frequently omitted in sequence analysis studies, although each has a relatively large impact on the genome. Here, we present GraphTyper2, which uses pangenome graphs to genotype SVs and small variants using short-reads. Comparison to the syndip benchmark dataset shows that our SV genotyping is sensitive and variant segregation in families demonstrates the accuracy of our approach. We demonstrate that incorporating public assembly data into our pipeline greatly improves sensitivity, particularly for large insertions. We validate 6,812 SVs on average per genome using long-read data of 41 Icelanders. We show that GraphTyper2 can simultaneously genotype tens of thousands of whole-genomes by characterizing 60 million small variants and half a million SVs in 49,962 Icelanders, including 80 thousand SVs with high-confidence.

The symposium ‘‘Genomics and Vertebrate Adaptive Radiation: A Celebration of the First Cichlid Genome’’ was held at the annual meeting for the Society of Integrative and Comparative Biology in Boston in January, 2009. While the central topic was the genomics of adaptive radiation in cichlid fishes, the symposium integrated speakers studying genotype-to-phenotype maps in other model organisms. Genomic studies investigating cichlid phenotypes were paralleled with studies using similar techniques to answer questions in other adaptive radiations. The topics were not restricted to any one research area in order to foster interdisciplinary discussions from developmental, physiological, and behavioral research with the common goal to apply and enhance genome-level analyses. Research programs such as the genetics and development of craniofacial diversity, behavior, and color patterns in zebrafish, mouse, and stickleback were described and then augmented with speakers focusing on the genetic and developmental architecture of cichlids’ jaws, behavior, and color. This integration across both established model systems and the new cichlid model sought to capitalize upon a comparative organismal framework that focused on how linking genomic information to studies of phenotype could hasten scientific discovery. The comparative and integrative approach inherent in most cichlid research extends to the current genome sequencing effort. Multiple cichlids have been chosen for genome sequencing because of their important applied as well as scientific value. The genomes of the cichlid fish Tilapia nilotica and as many as three other species representing the faunas of the largest East African Rift lakes will be sequenced. Astatotilapia burtoni could represent the 400 species present within Lake Tanganyika and its surrounding rivers because it is a genomic model for social regulation of behavior (Renn et al. 2008). Paralabidochromis chilotes could provide a representative for the !350 species from Lake Victoria, a radiation being strongly affected by ecological change (Seehausen et al. 2008). Metriaclima zebra could also be sequenced because it has been used to build a haplochromine linkage map (Albertson et al. 2005) and is a member of the !500–1000 species in the very recent radiation from Lake Malawi. These three other cichlid species, representing independent radiations within the East African Rift Lakes, span the incredible craniofacial, color, and behavioral diversity of the nearly 2000 species that inhabit this region (Salzburger 2008). Tilapia nilotica will be the first cichlid genome completed, and it is not only a major source of protein for much of the world but also is a focal organism for comparative renal physiology, ion balance, and endocrinology (Rengmark et al. 2007; Inokuchi et al. 2009). The sequencing of these cichlid genomes will provide an unprecedented view simultaneously into the genetics of one of the world’s major protein sources and into the unequalled phenotypic diversity of these fishes. In the symposium, researchers studying cichlids and other model vertebrate organisms focused on exploring the ways in which genomic studies can inform our mechanistic understanding of adaptive phenotypic diversity. The repeated phenotypic convergence in cichlid fishes makes this diverse vertebrate group exceptional for comparative analyses of genetics and development. Dr Darrin Hulsey described how cichlids are an ideal group to examine the similarities and differences in genomic and developmental changes that have lead to repeated convergence in neuralcrest-derived attributes as diverse as jaws, color,

It has recently been shown that structural variants (SV) can have a higher impact on gene expression variation compared to single nucleotide variants (SNV) in different plant species. Additionally, SV were associated with phenotypic variation in several crops. However, compared to the established SV detection based on short-read sequencing, less approaches were described for linked-read based SV calling. We therefore evaluated the performance of six linked-read SV callers compared to an established short-read SV caller based on simulated linked-reads in tetraploid potato. The objectives of our study were to i) compare the performance of SV callers based on linked-read sequencing to short-read sequencing, ii) examine the influence of SV type, SV length, haplotype incidence (HI), as well as sequencing coverage on the SV calling performance in the tetraploid potato genome, and iii) evaluate the accuracy of detecting insertions by linked-read compared to short-read sequencing. We observed high break point resolutions (BPR) detecting short SV and slightly lower BPR for large SV. Our observations highlighted the importance of short-read signals provided by Manta and LinkedSV to detect short SV. Manta and NAIBR performed well for detecting larger deletions, inversions, and duplications. Detected large SV were weakly influenced by the HI. Furthermore, we illustrated that large insertions can be assembled by Novel-X. Our results suggest the usage of the short-read and linked-read SV callers Manta, NAIBR, LinkedSV, and Novel-X based on at least 90x linked-read sequencing coverage to ensure the detection of a broad range of SV in the tetraploid potato genome.

Structural variants (SVs) are a largely unstudied feature of plant genome evolution, despite the fact that SVs contribute substantially to phenotypes. In this study, we discovered SVs across a population sample of 347 high-coverage, resequenced genomes of Asian rice (Oryza sativa) and its wild ancestor (O. rufipogon). In addition to this short-read data set, we also inferred SVs from whole-genome assemblies and long-read data. Comparisons among data sets revealed different features of genome variability. For example, genome alignment identified a large (∼4.3 Mb) inversion in indica rice varieties relative to japonica varieties, and long-read analyses suggest that ∼9% of genes from the outgroup (O. longistaminata) are hemizygous. We focused, however, on the resequencing sample to investigate the population genomics of SVs. Clustering analyses with SVs recapitulated the rice cultivar groups that were also inferred from SNPs. However, the site-frequency spectrum of each SV type—which included inversions, duplications, deletions, translocations, and mobile element insertions—was skewed toward lower frequency variants than synonymous SNPs, suggesting that SVs may be predominantly deleterious. Among transposable elements, SINE and mariner insertions were found at especially low frequency. We also used SVs to study domestication by contrasting between rice and O. rufipogon. Cultivated genomes contained ∼25% more derived SVs and mobile element insertions than O. rufipogon, indicating that SVs contribute to the cost of domestication in rice. Peaks of SV divergence were enriched for known domestication genes, but we also detected hundreds of genes gained and lost during domestication, some of which were enriched for traits of agronomic interest.

BackgroundAlzheimer's disease (AD) exhibits significant genetic heterogeneity, with no single model fully explaining its inheritance pattern. Structural variants (SVs) may contribute to the missing heritability of AD, particularly in disease pathology and familial AD, which remain largely unexplored.MethodStructural variants (SVs) were analyzed in unrelated 1,162 individuals from the ROSMAP study (687 cognitively normal (CN), 475 AD patients), as well as 197 Non‐Hispanic White (NHW) families (926 individuals, 58.5% with AD) and 214 Caribbean Hispanic (CH) families (1,340 individuals, 59.17% with AD). Large insertions and deletions were identified using Manta, Absinthe, and MELT from whole‐genome sequencing, supplemented by Sniffles2 and external SV call sets. This approach generated a unified VCF file containing 45,251 insertions and 76,566 deletions. After filtering for missingness rate and allele counts, a total of 9,290 insertions and 9,529 deletions were retained for analysis. Genome‐wide SV analyses were performed with AD‐related metrics in ROSMAP, adjusting for age, sex, and principal components, and validated in an independent ADSP replication cohort comprising 29,055 unrelated individuals (17,110 CN and 11,945 AD patients).ResultWe identified nine SVs (one insertion and eight deletions) segregating in AD families. An insertion near PDCD10 in NHW families reached genome‐wide significance (p <5×10‐8) for AD risk in the replication cohort. Two deletions in KIF26B and SPHKAP were found in both CH and NHW families, while five appeared only in NHW and one in CH. All eight deletions showed genome‐wide significance for AD risk. SV GWAS on AD pathologies in ROSMAP revealed an insertion in INPP4B with the strongest association to global brain pathology (β=0.19; p = 2.32×10‐5), also linked to pathological AD, diffuse plaques, and Braak staging. A deletion near ADGRB1 (β=1.53; p = 5.05×10‐6) was associated with neuritic plaques, and another near ARMC2 (β=1.66; p = 2.17×10‐6) was linked to diffuse plaques. All three SVs associated with AD (p <5×10‐6) in replication.ConclusionOur findings reveal numerous SVs segregating in familial AD across both CH and NHW families, as well as additional SVs linked to sporadic AD. Prioritizing these SVs based on their potential effects on gene function and expression will provide deeper insight into their contributions to both familial and sporadic AD.

African cichlid fish form new species faster than any other vertebrates, with hundreds of species evolving within the last 2 million years in Lake Malawi and within the last 120,000 years in Lake Victoria. This rapid speciation makes cichlids good models for elucidating the genetic mechanisms behind biodiversity. Vision may play a key role in cichlid evolution, adapting them to forage for new foods or colonize new habitats. Vertebrate retinas have two groups of light-sensitive proteins called opsins: those in rod photoreceptors, which are sensitive to dim light, and those in cone photoreceptors, which are sensitive to color. Changes in the visual system could be due to differences either in the expression of opsin genes or in their DNA sequences. A Research Article in this issue of PLoS Biology by Christopher Hofmann and colleagues suggests that both mechanisms underlie changes in visual sensitivity in cichlids. To investigate differences in cichlids’ visual sensitivity, the researchers compared 56 types of fish from Lake Malawi, where the water is spectacularly clear, and 11 types of fish from Lake Victoria, where the water is turbid. First, they used messenger RNA to compare the expression of opsin genes in the Malawi and Victoria fish. Cichlids have one rod opsin and six cone opsins, which range in sensitivity from ultraviolet to red wavelengths (roughly 300 to 700 nanometers). In addition, cichlids have two kinds of cones: single cones, which contain the three opsins that are sensitive to shorter wavelengths (ultraviolet, violet, and blue); and double cones, which contain the three opsins that are sensitive to longer wavelengths (blue-green, green, and red). The results showed that Malawi cichlids collectively expressed all six cone opsins and so were sensitive to wavelengths across the spectrum. The researchers then used opsin expression patterns in Malawi cichlids to estimate the sensitivities of their single cones, which contain the shorter wavelength opsins, and double cones, which contain the longer wavelength opsins. They found that Malawi cichlids clustered into three distinct groups that are sensitive to short, middle, and long wavelengths. This opsin diversity is likely driven by foraging. For example, Malawi cichlids that eat other fish generally expressed more long-wavelength opsins, while those that eat plankton and algae generally expressed more short-wavelength opsins. Ultraviolet-sensitivity is known to increase fishes’ foraging efficiency on zooplankton and other microorganisms, suggesting that these differences in opsin expression may be adaptive. In contrast, while all of the Victoria cichlids expressed the red-sensitive opsin, none of them expressed the ultravioletsensitive opsin. This makes sense because little ultraviolet light penetrates turbid water. However, expression of the violetsensitive opsin varied amongst Victoria cichlids. The diversity of this opsin is likely driven by light levels, which vary by nearly four orders of magnitude in Lake Victori-

Objectives and Rationale: Structural variants (SV) including large copy number aberrations (CNV), translocations, inversions, and large insertions and deletions (INDEL) play a critical role in tumorigenesis and progression. In fact, we now know that tumors can be categorized according to the size of mutations harbored. In several cancers, including ovarian and breast cancer, the large structural mutations, rather than single site mutations, play a dominate role in tumor etiology. Therefore, it's critical to implement reliable algorithm for the detection of structural variants in DNA sequencing data. Mate-pair sequencing is a protocol specifically implemented for detection of the whole-genome level structural variants. It requires less sequencing depth therefor is cost effective, and enables the detection of CNVs, translocations, and inversions simultaneously. However, so far there has been no reliable bioinformatics pipeline for the analyses of the mate-pair sequencing data. Methods: Our novel algorithm, the SnowShoes-SV, is an exhaustive search algorithm designed specifically for mate-pair DNA sequencing data analyses. It calls the SVs based on disconcordant read pairs. The false SVs are filtered according to the following criteria: (i) the number of the supporting read pairs; (ii) the lack of reads from control data that implicate SV at the same region; (iii) the mappability and uniqueness of the region based on data from the ENCODE project; (iv) consistencies of the mapping orientations of the supporting read pairs; (v) the similar sizes between sequencing library and the two end read clusters. Results: Using a set of samples previously genotyped by aCGH, the SnowShoes-SV successfully detected all known CNVs and other SVs. It also identified copy number neutral translocations and inversions previously not identified by aCGH. In addition, the algorithm nominated novel SVs which are to be validated by PCR. Conclusions: SnowShoes-SV is a highly sensitive and specific algorithm for SV detection from the mate-pair DNA sequencing data. Citation Format: Yan W Asmann, Chen Wang, Brian M Necela, Xianfeng Chen, Jean-Pierre A Kocher, Matthew J Maurer, Thomas M Habermann, Susan L Slager, Andrew L Feldman, Anne J Novak, James R Cerhan, Edith A Perez, E Aubrey Thompson. An exhaustive algorithm for detecting copy number aberrations and large structural variants in whole-genome mate-pair sequencing data [abstract]. In: Proceedings of the Thirty-Seventh Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2014 Dec 9-13; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2015;75(9 Suppl):Abstract nr P6-07-03.

Structural variants (SVs) are increasingly recognized as key contributors to human diseases. However, our understanding of SVs in health and disease is limited, mainly due to their structural complexity and variable length in individuals, as well as limitations inherent to the available genomic technologies and reference genome used. To systematically evaluate SVs across human whole-genome samples using hg38/GRCh38 and gapless T2T-CHM13 references, we introduced an innovative multiplatform approach, LongReadChecker (LoReC), which advances SV comparison and annotation based on distance variance, intersection, gene overlap, and the closest SV in the clinical database. Comparison of the performance in detecting SVs from public and our own whole-genome datasets from short-read sequencing (SRS), available long-read sequencing (LRS) platforms, and optical genome mapping (OGM) revealed that most SVs detected by SRS were confirmed by LRS, but LRS can identify twice as many SVs (25,000 SVs/genome) with greater read mapping accuracy. Our LongReadChecker (LoReC) analysis further highlights the utility of the T2T-CHM13 reference in SV detection, as 20% more deletions and 20% less insertions were detected compared with hg38/GRCh38, which was particularly evident in long-read datasets. Since 80% of the SVs detected by LRS/SRS are smaller than 0.5 kbp, OGM did not detect them. Our study revealed that introducing distance variance, intersection, gene overlap, and the closest SV in the clinical database may help compare and annotate SVs in diagnostics. Our data showed that LRS, together with T2T-CHM13 gapless sequences, can improve the diagnostics of patients with human diseases when SRS fails to identify the cause.

Peering into the gaps: Long-read sequencing illuminates structural variants and genomic evolution in the Australasian snapper

Pangenome can reveal a large number of variations, providing a more comprehensive view of the genetic diversity of species that a single reference genome cannot surpass. Here, we assembled the haplotype telomere-to-telomere genome and 10 chromosome-level genomes, integrated with two previously reported genomes, and constructed a graph pangenome for P. mume with 'PmNH_Hap1' T2T genome as the reference. Core gene families accounted for 64.47%, while non-essential gene families comprised 35.53%. The total length of the graph pangenome reaches 412.41 Mb, which is a significant increase of 179.60 Mb compared with the 'PmNH_Hap1' reference genome, and these unique insertion sequences contain 5918 genes. Additionally, by combining with five Prunus armeniaca genomes for pangenome variation detection, we identified a total of 51 461 non-redundant SVs, including 15 217 insertions, 21 683 deletions, 13 509 translocations and 1025 inversions. Among these, 60.50% of presence/absence variations were formed by transposons. Pangenome analysis revealed that a 376 bp SV insertion was universally present in the promoter of this gene in P. mume, PmAGL30 can bind to this SV to enhance PmPH4 gene expression and promoted citric acid accumulation in P. mume fruits, leading to extremely significant differences in citric acid content compared with other stone fruits. Additionally, we developed SV molecular markers for the early screening of germplasm with high citric acid content in P. mume fruits. In summary, we constructed a high-quality graph pangenome that reveals abundant genetic variations, providing valuable insights for variety improvement and molecular breeding in P. mume.

Here we describe the ways in which the sequence and annotation of thePlasmodium falciparum reference genome has changed since its publication in 2002. As the malaria species responsible for the most deaths worldwide, the richness of annotation and accuracy of the sequence are important resources for theP. falciparum research community as well as the basis for interpreting the genomes of subsequently sequenced species. At the time of publication in 2002 over 60% of predicted genes had unknown functions. As of March 2019, this number has been significantly decreased to 33%. The reduction is due to the inclusion of genes that were subsequently characterised experimentally and genes with significant similarity to others with known functions. In addition, the structural annotation of genes has been significantly refined; 27% of gene structures have been changed since 2002, comprising changes in exon-intron boundaries, addition or deletion of exons and the addition or deletion of genes. The sequence has also undergone significant improvements. In addition to the correction of a large number of single-base and insertion or deletion errors, a major miss-assembly between the subtelomeres of chromosome 7 and 8 has been corrected. As the number of sequenced isolates continues to grow rapidly, a single reference genome will not be an adequate basis for interpreting intra-species sequence diversity. We therefore describe in this publication a population reference genome ofP. falciparum, called Pfref1. This reference will enable the community to map to regions that are not present in the current assembly.P. falciparum 3D7 will continue to be maintained, with ongoing curation ensuring continual improvements in annotation quality.

Does hybridization play a broad innovative role in evolution? Many studies have shown hybrid origins of individual species, particularly in major adaptive radiations, but this may be a consequence, rather than a cause, of the existence of many closely related species. Cases of hybridization in the early stages of major adaptive radiations are comparatively rare. Here, we report phylogenetic evidence for ancient introgression between distinct lineages of the species-rich Lake Malawi haplochromine cichlid fishes. Mitochondrial DNA (mtDNA) sequences indicated surprisingly close relationships between the shallow-water rocky habitat "Mbuna" species and a group of dark-adapted "Deep-Benthic" species specialized for feeding in low-light conditions (dawn/dusk, under overhangs, and deep water). By contrast, analyses of nuclear amplified fragment length polymorphism data demonstrated that these Deep-Benthic cichlids were more closely related to shallow water "Shallow-Benthic" soft-sediment feeders, a group that shares similar head and body morphology. A coalescent-based computer simulation indicated that the mtDNA similarity of rocky habitat Mbuna species and dark-adapted Deep-Benthic species was due to hybridization rather than incomplete lineage sorting. Comparisons of morphology indicated that some Deep-Benthic species possessed novel morphology not present in other Lake Malawi species groups. Thus, these analyses support the hypothesis that ancient hybridization occurred within the Lake Malawi cichlid radiation, that the event occurred before the radiation of a species group adapted to low-light benthic habitats, and that this group went on to dominate the deep-water regions of Lake Malawi. The results of this study contribute to a growing literature consistent with a creative role of hybridization in the evolution of species diversity and adaptive radiations.