Airport experiment sequencing cost estimate

We are planning to sequence samples from four sources (redacted because this is publicly visible). We will receive one sample from each source each day weekday for eight weeks. Thus we will have \(8 \times 5 \times 4 = 160\) samples total.

We will the following steps ourselves:

• Concentrate the samples
• Filter out large particles (to enrich for viruses)
• Extract RNA

Then we will deliver the RNA extracts to the BioMicroCenter for:

• RNA seq library prep:

  • Standard RNA prep (not high-throughput), pricing and QC are per sample
  • Total RNA sequencing, includes sample QC, ribosomal depletion, generation of cDNA, library construction, indexing and library quality control

• Sequencing on an Illumina NovaSeq 6000 S4 with 300nt per read pair.

The goal of this doc is to estimate the cost of library prep and sequencing.

Sources

• The cost of library prep and sequencing come from the BioMicroCenter’s pricing page. We use the MIT prices.
• Characteristics of the NovaSeq come from illumina (Table 1).

Estimate

Each NovaSeq flow cell has 4 lanes. When run in 2 x 150 bp mode, it generates 150 basepair forward and reverse read pairs for 300 bp per read pair. In one run, the flow cell generates 2400–3000 Gb of data. In the following we will make a conservative estimate by using the lower end of the range.

lanes_per_flow_cell = 4
bp_per_read_pair = 300
gb_per_flow_cell = 2400

read_pairs_per_flow_cell = gb_per_flow_cell * 1e9 / bp_per_read_pair
read_pairs_per_lane = read_pairs_per_flow_cell / lanes_per_flow_cell
print(f"We expect at least {read_pairs_per_lane/1e6} M read pairs per lane")

We expect at least 2000.0 M read pairs per lane

The smallest unit of NovaSeq sequencing we can buy is one lane, which costs $5,940. We also pay $258.5 per sample for library preparation.

cost_per_lane = 5940
library_cost_per_sample = 258.5

We’ll consider three options:

1. Sequencing all 160 samples in 4 lanes
2. Sequencing all 160 samples in 8 lanes
3. Sequencing MWF only (\(160 \times 3/5 = 96\) samples) in 5 lanes

for num_samples, num_lanes in [(160, 4), (160, 8), (96, 5)]:
    samples_per_lane = num_samples / num_lanes
    read_pairs_per_sample = read_pairs_per_lane / samples_per_lane
    sequencing_cost = cost_per_lane * num_lanes
    library_cost = library_cost_per_sample * num_samples
    total_cost = sequencing_cost + library_cost
    cost_per_sample = total_cost / num_samples
    cost_per_read_pair = cost_per_sample / read_pairs_per_sample
    print(f"""
    {num_samples} samples in {num_lanes} lanes:
        Million read pairs per sample:\t{read_pairs_per_sample / 1e6:3,.0f}
        Sequencing cost:\t${sequencing_cost:7,.0f}
        Library prep cost:\t${library_cost:7,.0f}
        Total cost:\t\t\t${total_cost:7,.0f}
        Cost per sample:\t${cost_per_sample:7,.0f}
        Cost per M rp:\t\t${cost_per_read_pair * 1e6:10.2f}
    """)

    160 samples in 4 lanes:
        Million read pairs per sample:   50
        Sequencing cost:    $ 23,760
        Library prep cost:  $ 41,360
        Total cost:         $ 65,120
        Cost per sample:    $    407
        Cost per M rp:      $      8.14
    

    160 samples in 8 lanes:
        Million read pairs per sample:  100
        Sequencing cost:    $ 47,520
        Library prep cost:  $ 41,360
        Total cost:         $ 88,880
        Cost per sample:    $    556
        Cost per M rp:      $      5.56
    

    96 samples in 5 lanes:
        Million read pairs per sample:  104
        Sequencing cost:    $ 29,700
        Library prep cost:  $ 24,816
        Total cost:         $ 54,516
        Cost per sample:    $    568
        Cost per M rp:      $      5.45