Micro Mirror Module analysis of small seed samples for soybean seed quality

S.L. Naeve (1), J. Orf (1), A. Killam (1), W. Shadow (2), and D. Honigs (2)

(1) University of Minnesota, 411 Borlaug Hall, St. Paul, MN 55108 USA, email: naeve002@umn.edu
(2) Perten Instruments Inc., 6444 S. Sixth Road, Springfield, IL 62712 USA

This article is adapted from a paper presented at 14th International Conference on Near Infrared Spectroscopy, 7–13 November 2009, Bangkok, Thailand.

Soybean seed is highly prized worldwide for its valuable protein and oil. A mirrored sample cup has been used for NIR analysis of very small seed lots (8–14 seeds) to quickly screen germplasm for appropriate quality traits.

Breeders need to quickly and cheaply evaluate breeding lines and other small samples for quality characteristics. Individual soybean plants may hold only 100–200 seeds and only a limited number are available for analysis because whole seed must be retained for planting. In addition, thousands of breeding lines must be evaluated annually.

A Micro Mirror Module can be used to evaluate a small number of whole soybean seeds for common quality parameters with a level of precision that approaches that of traditional techniques utilizing large samples. Multiple sub-samples may improve the level of precision in certain circumstances.

Materials And Methods

Five unique seed lots (M1–M5) which differed in protein, oil, and linolenic acid concentrations, seed size, and pedigree, were used:

• M1: MN1806SP, large seed, high protein, 388 seeds 100g-1, 258 mg seed-1
• M2: MN1702SP, low linolenic oil, 673 seeds 100g-1, 149 mg seed-1
• M3: M129, non-nodulating, 782 seeds 100g-1, 128 mg seed-1
• M4: MN1410, normal seed size, 516 seeds 100g-1, 194 mg seed-1
• M5: MN1203SP, small seed size, 851 seeds 100g-1, 118 mg seed-1

Approximately 90 g of each seed lot was used for protein and oil analysis by near infrared spectroscopy using a Perten DA 7200 Feed Analyzer with a standard small metal cup (75 mm diameter). Each sample was scanned and repacked 20 times to provide average reference quality profiles (Table 1).

Samples were divided into 12–15 sub-samples (8 to 14 seeds) for analysis using the Micro Mirror Module. Each sub-sample was scanned and repacked 15 times (Table 2).

Data were analyzed using the GLM and MIXED procedures of SAS v9.11. Scan to scan error of the 90 g sample in the standard small cup, sampling error as estimated by the 12–15 sub-samples per seed lot, and scan to scan error through repacks and scans of the same 8–14 seed sub-samples were investigated (Table 3).

Minimum number of samples and scans required for the Micro Mirror Module to meet the precision of the standard small cup were calculated based on two criteria: combinations of samples/scans of each sample that resulted in the lowest number of total scans, and combinations of samples/scans that resulted in the lowest number of samples required (Table 4).

Results

Table 1. Mean and standard deviation values for nine soybean seed constituents

Table 2. Mean and standard deviation values for nine soybean seed constituents; sub-samples of 8–14 seeds each.
Table 3. Percent of total variance attributed to either sub-sampling error or scan error within five soybean seed lots or percent of total variance attributed to the sample itself, sub-samples within samples, or scans within sub-samples across five soybean seed lots scanned using the Micro Mirror Module.


Table 4. Number of samples and scans of each sample required to meet confidence intervals based on 95% probability of accuracy within provided ranges for selected soybean constituents for both the standard small cup and the Micro Mirror Module.

Note: ‘-‘ is used to identify samples where more than 10 samples are required to meet the precision of the small cup.

Conclusions

  • The Perten Micro Mirror Module provides confidence interval of about 2x that of the standard small cup.
  • There is significant error between 8–14 seed sub-samples due, in large part, to heterogeneity within seed lots.
  • For well-characterized constituents (protein and oil), total error variance for sample analysis with the Micro Mirror Module was 4% or less. Ninety-six percent or more of the error variance for these constituents was due to variation between seed lots.
  • Sampling error is the major source of error when using a small number of seeds. This error can be overcome by analysis of multiple sub-samples from the same seed lot.
  • For protein estimation, 2–4 samples are required to meet a +1% confidence interval when using the Micro Mirror Module. For oil, a +1% confidence interval can be met with only one or two samples.
  • We have demonstrated that the Micro Mirror Module can be used to evaluate a very small number of whole soybean seeds for common quality parameters with a level of precision that nears that of traditional techniques utilizing large samples. However, we determined that multiple sub-samples may be required to achieve a necessary level of precision in certain circumstances.

References

1, SAS Institute (2002). The SAS system for Windows. 9.1, SAS Inst., Cary, NC.

2, R.A. Illipronti Jr., Lommen, W.J.M, Langerak, C.J. and Struik, P.C. (2000). Netherlands J. of Agric. Sci. 48, pp. 165–180.

3, F.I. Collins and Carter, J.L. (1956). Agron. J. 48, pp. 216–219.

4, J.B. Carlson and Lersten, N.R. (2004). In H.R. Boerma and J.E. Specht (eds), Soybeans: improvement, production, and uses,  3rd ed., ASA and SSSA, Madison, WI, pp.59–95.