Intro to Cancer Genomics

Hallmarks of Cancer

• tumor cells acquire abnormal abilities by co-opting normal cell behavior

Karyotyping of Colorectal Cancer Cell Lines

• numerical and structural chromosomal instability

• translocation b/w chromosomes

• extra chromosomes (genome may be doubled)

Cancer is a Genetic Disease

• genomic and epigenomic alterations

• somatic copy number alternations (SCNA)

Cancer Cells accumulate somatic alterations over time

• chemotherapy might induce mutations over time

• mutations might break down cell functions

• mutations in specific places might cause cancer

Mutation Burden

• mutation burden varies by cancer type, exposure, age of onset, DNA repair ability, etc.

• some cancer types have more mutations (e.g. exposure to UV/smoking)

  • pediatric cancer have less mutations, but alterations end up being more disruptive

Cancer Genomics Pipeline

• comparing to person’s own blood instead of reference blood (for detecting somatic mutations)

Cancer Genomics Pipeline: nf-core/sarek

• nf-core/sarek is a workflow designed to detect variants on genome sequence data

• can work on any species w/ a reference genome

• can handle tumour/normal pairs

• built using nextflow (a workflow tool)

  • uses Docker/Singularity containers making installation trivial and results highly reproducible

nf-core/sarek Overall Workflow

• input raw sequencing files (FASTQ) or pre-aligned BAM files and reference genome

• Align reads to the reference genome

• Sort BAM files and mark duplicates

• Call variants: identifying positions in a genome where a sample’s DNA sequence differs from the reference genome

• Variant Filtering: Remove low-confidence variants

• Annotation: translates raw genetic differences into biologically and clinically interpretable information.

Calling of Single Nucleotide Variants

• germline mutations are also detected (we ignore); only looking at somatic mutations

• sequence both samples (tumor +normal DNA)

• align reads to reference genome (generate BAM files for tumor and normal)

• compare at each genomic position

  • tumor = variant, normal = no variant → somatic mutation

  • tumor = variant, normal = variant → germline variant

• statistical modeling: evaluating read depth, variant allele fraction, base quality, tumor purity

  • estimating the probability that the variant exists only in tumor and not sequencing noise

Cancer Genome Variation: Sequencing Read Alignments

• multiple types of cancer genome variation may be inferred from sequencing read alignments

Sample Purity on Coverage

• if a sample was completely pure, variants are detectable at low coverage (don’t have to cover genome very deeply)

  • every read at a variant site comes from a cell that actually has the mutation

  • even a low number of sequencing reads can reliably detect the variant

• variant allele fraction is higher in pure samples: heterozygous mutation → 50% of reads show the variant; homozygous mutation →100% of reads show the variant

  • in mixed samples (tumor+normal), normal cells dilute the signal and the variant allele fraction drops → requires higher coverage to confidently detect variants

Are tumor samples pure?

No! In reality, tumors are a mix of cancer and normal cells

• in cancer genomics, we need to consider “tumour content” or “sample purity”

• tumour purity (or lack of) makes calling cancer mutations difficult

• as such, sequencing a cancer genome requires sufficiently deep coverage, especially for samples w/ low tumour content

  • increases the chances that alternate alleles (mutations) are detected

Somatic Copy Number Alteration (SCNAs)

• prevalent, acquired genomic changes in tumor cells (not inherited) involving the gain (amplification) or loss (deletion) of DNA, ranging from small segments to entire chromosome arms

  • changes relative to one’s ploidy (need to determine current ploidy state before determining SCNAs)

• major drivers of cancer development, progression, and heterogeneity, affecting oncogene and tumor suppressor gene dosage

• Compare tumor vs. normal at the same locus. Gain (amplification) → tumor has more copies than normal. Loss (deletion) → tumor has fewer copies than normal.

• Detection: higher coverage than expected → gain, lower coverage than expected → loss

Calling Somatic Copy Number Alterations

• most SCNA callers use a read-depth based approach (focus on SNPs)

• two main input channels:

  • Log₂ ratio (logR): relative depth between tumor and normal

  • B-allele fraction (BAF): allelic imbalance, gives the ability to call allele-specific copy number

  • Combining BAF with logR allows you to see:

    • Which allele is lost or amplified

    • If there is copy-neutral LOH (no change in logR, but BAF shows imbalance)

• data are segmented into regions of constant copy number (i.e blue lines)

• segments are classified into copy number events

Log₂ Ratio Figure

• data from this plot is sufficient to say whether there is a change in ploidy (gain or loss)

• compares sequencing coverage in the tumor to the normal at each locus

• logR > 0 → tumor has more DNA than normal → gain/amplification

• logR < 0 → tumor has less DNA than normal → loss/deletion

• logR ≈ 0 → no change (copy number = normal)

B-allele fraction Figure

• fraction of reads supporting the alternate allele at heterozygous SNPs.

  • see if representation is equall b/w alleles

• Normally, for a germline heterozygous SNP: BAF ≈ 0.5 (50% reference, 50% alternate)

• allelic imbalance: Copy number changes can shift the balance between alleles.

  • Loss of one allele (LOH) → BAF → 0 or 1

  • Gain of one allele → BAF shifts toward 0.33 or 0.66

Neutral Loss of Heterozygosity

• shift in allele representation, but no visible gain or loss

• occurs when one allele is lost in replaced by other allele (still the same amount of copies, but there’s a loss in heterozygosity)

  • no net gain

• wouldn’t show as a change in logR, but in BAF, deviates from 0.5 (allelic imbalance)

  • Instead of a single band at 0.5, heterozygous SNPs split toward 0 and 1, forming two “allele-specific” clusters

Tumor Purity on SCNA Signal

• tumor purity affects SCNA signal

• lower purity means fewer cells harbor the SCNA events (harder to see SCNAs)

• weaker signal (signal to noise ratio is decreased)

Translocations: Soft-Clipped Bases

• A translocation occurs when a segment of DNA from one chromosome is moved and attached to another chromosome.

• When reads are aligned to the reference genome:

  • Sometimes only part of a read aligns and the remaining portion does not match the reference at that location.

  • That unmatched portion is called soft-clipped.

• translocations can be detected from soft-clipped bases

  • Soft-clipped bases can represent sequence that belongs to a different genomic location — potentially another chromosome.

Translocations: Soft-Clipped Bases FIGURE

Rearrangement Complex

• rearrangements can be highly complex and detectable at a base pair resolution

• genomic rearrangements include: translocation, inversions, deletions, etc.

• in cancer, rearrangements can involve multiple chromosomes, fragmented DNA segments, etc.

• tumors can show multiple breakpoints close together, chains of rearrangements, regions shattered and stitched back tgt, copy number changes intertwined w/ structural variants

• With high-throughput sequencing: Split reads can pinpoint the exact nucleotide where DNA breaks and rejoins.

  • We can identify the precise breakpoint sequence.