Nanopore sequencing at Mars, Europa, and microgravity conditions

Parabolic flight

Flight operations were conducted on November 17, 2017 onboard a Boeing 727-200F aircraft (G-Force One®, Zero Gravity Corporation). Four sets of parabolas were performed with 5, 6, 4, and 5 parabolas respectively (Fig. 1a). The first set targeted, in order, Mars g, Mars g, Lunar g, 0 g, and 0 g (Fig. 1b). All other parabolas targeted 0 g. The flight profile was segmented into periods of “transition”, “parabola”, “hypergravity”, and “other” (typically, gentle climb, descent, straight and level flight, or standard rate turns) on the basis of accelerometer measurements4. Sequencing was also performed on the ground prior to the flight as a control.

Fig. 1: Single molecule sequencing during parabolic flight.

a Phases of flight timeline (black: other/1 g; red: hypergravity; magenta: transition; blue: parabola). b Phases of flight for first set of parabolas. c Vibration (blue line, left axis) and sequencing reads measured during flight; each read is represented by a horizontal line (mux = black, run = red) at its representative read quality score, (q_{bar p}).


Sequencing of control lambda deoxyribonucleic acid (DNA) was performed for a total of 38 min on the ground and 103 min during flight, on the same flow cell, resulting in 5293 and 18,233 reads for ground (Supplementary Fig. 1) and flight (Fig. 1c; Supplementary Fig. 1) respectively, of which 5257 and 18,188 were basecalled (Supplementary Tables 1, 2). Of the flight reads, 14,431 fell wholly within a phase of flight, including parabola (404), hypergravity (1996), transition (7), and other (12,024). Sequencing reads were obtained during all parabolas, including under Mars, lunar/Europa, and zero-g conditions (Fig. 2). The g levels achieved during each parabola were previously reported4. For the purposes of statistical analysis, mux reads (Fig. 1c, black horizontal lines) were excluded to avoid any sequencer start-up effects.

Fig. 2: Sequencing in reduced gravity.

a g level achieved (black line) and RMS vibration (1 s bins, blue line) and associated sequencing reads acquired during first set of parabolas. Each read is represented by a horizontal line (gray: partially or completely in transition period; red: completely in non-transition period) at its representative read quality score, (q_{bar p}) (right axis). Vertical gray bands demarcate transitions between phases of flight. b Top scoring BLAST results for highest quality “Mars” read, indicated via arrow in a, length 6402. c Start of best match sequence alignment, to J02459.1 Enterobacteria phage lambda, complete genome, length 48502 (range 20562–27113, score 8907 bits(9877), expect 0.0, identities 6108/6651 (92%), gaps 395/6651 (5%), strand Plus/Minus). d Average genomic coverage of lambda for all parabolas based on tombo-aligned bases.


Zero-phase filtering effectively removed frequencies at or below 10 Hz (Supplementary Figs 26). Filtered root-mean-square (RMS) vibration varied throughout the flight and showed clear deviations associated with parabolas (Figs 1c, 2a; Supplementary Fig. 2), indicating a smoother environment during freefall. Remaining aircraft-associated vibrations were largely in the 10 Hz to 1 kHz band with peaks at 116–128, 250–270, 495–496, 580–680, and 876 Hz (Supplementary Fig. 6). During zero-g parabolas, the magnitude of the residual g level and vibrations were comparable (Fig. 2a).

Integrated read-level analysis

Stepwise linear regression was used to determine whether time and RMS vibration could predict median sequence quality (Supplementary Fig. 1), the Phred quality score5,6 associated with the average per-base error probability of a given read (see Materials and Methods). Unlike ground operations, where time was the only significant predictor of sequence read quality (p = 0), time, g level, and their combined effects were predicted to be significant indicators during flight (all p < 10−4; Supplementary Tables 3, 4). However, in both cases, the variance explained was small (adj. R2 = 0.060 and 0.275, respectively, for ground and flight).

In order to elucidate the role of g level on read quality, those reads falling wholly within an individual phase of flight were examined using a one-way ANOVA, with Tukey’s Honest Significant Difference (HSD) post hoc analyses (Supplementary Table 5). Sequence quality was significantly different during each phase of flight, with the lowest read quality during parabolas ((q_{bar p}) = 8.3) and the highest quality ((q_{bar p}) = 8.7) during hypergravity (Supplementary Fig. 7).

Integrated base-level analysis

Tombo7 was used to associate raw ionic current signals with specific genomic bases, and the number of reads aligning was similar to the number of reads with Phred quality scores5,6 >6.5. The percentage of bases that aligned to the lambda genome via tombo7 was 87.8% and 89.7% for ground and flight, respectively (Supplementary Table 1). Average coverage for tombo-aligned bases was adequate to sequence the lambda genome many times over during each parabola (Fig. 2d) and the coverage was largely explained by parabola duration (adj. R2 = 0.807; Supplementary Table 6). Ionic current levels associated with unique subsequences (k-mers) were similar between ground and flight conditions (Supplementary Fig. 8).

By aligning ionic current signals to bases, tombo allowed us to measure the translocation time associated with each base (Supplementary Fig. 9), the time required for the motor protein, acting as a ratchet, to move the DNA strand one base into the nanopore. Translocation here refers to motion of the motor protein relative to the DNA strand, and not the total time to get through the nanopore, which requires many translocation steps. The inverse of translocation time is a direct measurement of sequencing rate for a given nanopore (bases/s).

Translocation times were similar, but statistically longer, during flight as compared to ground. Despite the nearly sixfold (5.89) average higher RMS vibration during flight compared to ground (Fig. 3a), the probability densities for translocation time are strikingly similar (Fig. 3b; Supplementary Fig. 9). However, base translocation times were significantly different (Kolmogorov–Smirnov, two-tailed, p = 0, test statistic 0.0306), with a slight shift toward longer translocation times during flight. Notably, the median base translocation times were identical (seven samples or 1.8 ms) and the means only differed by 0.125 ms (2.2786 ms, ground; 2.4035 ms, flight). Thus, translocation times were robust to large variations in vibration.

Fig. 3: Translocation time is weakly or not affected by vibration.

a RMS vibration distributions for ground and flight. b Nanopore translocation time as measured by alignment of ionic current to the genomic reference: distribution for <10 ms. Ground (blue), flight (light brown), both (dark brown).

Ionic current noise is the variation in the flow of ions passing through the nanopore, measured here at the per-base level as a normalized signal standard deviation determined by tombo through optimal alignment of measured ionic current to a genomic sequence7. A stepwise linear regression was performed to determine if time, RMS vibration, or their combined effects were significant predictors of ionic current noise during ground (Supplementary Fig. 10; Supplementary Table 7) and flight (Supplementary Fig. 10; Supplementary Table 8) operations. Flight analysis included the additional variable g level.

For ground operations, the impact of time alone was not significant. However, both vibration (p = 0.0018) and the interaction effect of time and vibration (p = 0.041) were significant predictors of ionic current noise (Supplementary Table 7). However, the explanatory power of the regression was low (adj. R2 = 0.009). Conversely, time was the only significant predictor of the effect on ionic current noise during flight. Neither RMS vibration, g level, nor any of their respective combined effects had significant impacts on ionic current noise (Fig. 4; Supplementary Table 8).

Fig. 4: RMS vibration and median ionic current noise during flight.

The single 1 s period with median ionic current noise >0.5 has a median absolute deviation (MAD) of >15 and is therefore an outlier (typically defined as MAD >3).

Because time was a significant indicator of ionic current noise during flight, it was necessary to assess whether the effect could be attributed to a specific phase of flight (Supplementary Table 9). Tukey’s HSD post hoc test demonstrated that out of all six possible pairwise comparisons, only one, parabola vs. transition, was not significant (two-sided p = 0.345; other p < 10−3). Ionic current was significantly lower in hypergravity, parabola, and transition phases as compared to other. Ionic currents during hypergravity phases were, on average, lower than all other phases (Supplementary Table 9; Supplementary Fig. 11). Thus, while the impact of phase of flight on read quality showed a trend toward higher read quality with higher g level (Supplementary Fig. 7), no such pattern was observed with ionic current (Supplementary Fig. 11).

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *