- Email: [email protected]
A report of an international collaborative experiment to demonstrate the uniformity obtainable using DNA profiling techniques
30
a series of samples using its own protocols (the only standardisation was the use of HinfI as the restriction enzyme and use of a common ladder marker). It was demonstrated that the sizes of fragments determined by different laboratories were within a match window of 10%. Although it was possible to compare directly results between laboratories using this wide window, interpretation and allocation of statistical significante was more difficult when such large differences were obtained between laboratories. The second series of experiments described in this paper was carried out in order to determine whether comparable results could be obtained if different laboratories used the same protocol. Clearly, it was not possible to standardise completely because different equipment was in use. However, it was possible to standardise on electrophoretic buffer and make of agarose. Also each laboratory was supplied with a full protocol and samples of DNA; equipment was not standardised. A comparison of the within- and between-laboratory measurement error has been evaluated. Materials and Methods
DNA was bulk-extracted by one laboratory (Centra1 Research & Support Establishment CRSE) from three different blood samples. Half was then restricted by this laboratory with 30 x excess Hinff before distribution and the other half was supplied to laboratories unrestricted so that the effect of different restriction methods could be compared. In addition, HinfI restricted K562 (genomic control; Promega) and Amersham lambda ladder was supplied to each laboratory. Each laboratory was supplied with a full protocol; ICN Seakem GTG agarose was distributed from the same bottle. The agarose (2.2 g) was added to 315 ml Tris-borate EDTA (TBE) (10 x TBE = 1.3 M Tris, 0.75 M boric acid, 25 mM ethylenediaminetetraacetic acid (EDTA); pH 8.8) to make a 20 x 25 cm gel. Gels were electrophoresed in TBE buffer at 55 V until the 2 kb marker of an ethidium bromide-saturated X HindIII marker had travelled 16 cm along the gel (otherwise the electrophoretic system was ethidium bromide-free). Southern blots were probed with YNH24 [2] and MS43a [3]; fragments were sized at CRSE using Biotrac (Foster and Freeman Ltd). The experimental design is illustrated in Fig. 1. Each laboratory followed the protocol supplied. Both unrestricted and restricted DNA was supplied by the coordinating laboratory. Each participating laboratory was asked to restrict DNA supplied using their usual method. Results and Discussion Differences between samples restricted in-heuse compared to those restricted in the participating laboratmy
Minor differences were observed between samples restricted by CRSE compared to samples restricted in the participating laboratory (Appendices 1 - 3).
4.2
3.5
2.7
Fig. 1. Photograph of an autoradiograph showing the experimental design. The left hand side of the autoradiograph consisted of samples restricted by CRSE whereas the right hand side consisted of samples restricted by the participating laboratory. h, ladder marker; g, genomic control (not used in this exercise); a, sample a; b, sample b; c, sample c; k, K56.2.
32
Band positions on each half of the gel for each laboratory were compared using a Wilcoxon sign rank test i.e. laboratory 1 CRSE restricted samples A, B, C, K562 (YNH24 and MS43a) were compared with laboratory 1 restricted samples A, B, C, K562 (YNH24 and MS43a). Remaining laboratories were tested separately. Results were usually significant: the CRSE samples were restricted at the same time (and in the same way) on both halves of the gel, but stil1 gave a significantly different result. The differente was probably systematic. The reason for the effect was almost certainly because of the linear interpolation method for sizing bands between control ladder markers. The schematic diagram in Fig. 2 shows a greatly exaggerated distortion (purely for illustrative purposes) in both halves of the electrophoretic plate. In this example, the control sample is displaced in opposite directions in the right hand side compared to the left hand side of the plate because of curvilinear electrophoretic distortion; hence the systematic band shifts are correlated. In the current experiment (Fig. l), the differences between fragment sizes from each half of the gel were low (< 0.5% kb); it is very unlikely that different restriction buffers or different manufacturer’s í%~fI restriction enzyme contribute to the effect.
-
-m
_---
-_
___d
--
_ _
&
-
-_
-.
-4
r--
s---w_
-4
--_
--___-4 _M---_ ti--@
---___d
_-----. -.
__-
&#
/-
F
-w-
‘-
Fig. 2. A schematic diagram showing reasons for correlated shift due to electrophoretic phenomenon. The linear interpolation method does not take account of curvilinear distortion in the gel, hence in the above example (greatly exaggerated) both bands are shifted towards the high molecular linear interpolation. weight region in the right hand side of the gel. (- - - -), distortion; (-),
33
Calculation of okviation frorn the interlaboratory mean Laboratory 12 showed significant deviation from the mean, particularly for high molecular weight fragments >9.3 kb. This was probably because electrophoresis was carried out on a particularly hot day when buffer temperature could have reached >3O”C. Results from this laboratory were excluded when calculating interlaboratory means and coefficients of variation. Even so, results obtained in laboratory 12 from the lower molecular weight fragments detected by YNH24 were comparable to those from other laboratories. In Tables 1 and 2, fragments from both YNH24 and MS43a are given. They are arranged in descending size order for each laboratory. Between-laboratory means (M> and coefficients of variation were calculated. For each fragment, deviation from the mean (ìï@ was estimated and the data were summarised to show average laboratory deviation. Results from each half of the autoradiograph were compared (i.e. one half contained samples restricted by CRSE, Table 1 and the other half contained samples restricted by the participating laboratory, Table 2). The between-laboratory coefficient of variation was greatest for high molecular weight fragments (lO- 14 kb); an overall average of 0.72% kb was recorded for the whole range in Table 1. The mean laboratory deviation was no greater than 0.8% kb; this compared with an approximate 5.15% kb (YNH24) variation observed in the first series of EDNAP experiments for one sample, number 12 [ll. There were no obvious differences seen when testing samples restricted by the participating laboratories as compared to samples restricted by CRSE (between-laboratory mean values (IM>were very similar and no greater than 0.2% kb different from each other). Some laboratories showed predominantly negative deviation from the mean (e.g. laboratories 3 and 12; Table 1) whereas others were positive (laboratory 2, Table 1); laboratories 1 and 13 showed positive deviation in Table 1 and negative deviation in the second half of the autoradiograph (Table 2); this was a reflection of correlation of fragment deviation dependent upon unpredictable minor electrophoretic differences between two halves of a gel. An analysis of data wing a variable match criterion Evett et al. [4] introduced a method of analysis which enabled a large number of comparisons to be carried out from a relatively smal1 database. The principle of the test involves comparison of each pair of band measurements with every other pair in the database referring to a match guideline. Part of the datafile of sample A (YNH24) is as follows: Identifier (laboratory)
Band Weights (bp) 1 2
1
4564 4584 4633
2 3
2724 2735 2776
1
2
no.
3
4
4607 4001 3245 2917 2875 2762
4756
6
2901 2857 2757 1958
14249 10607 10270 9019 5312 5225 4808 4717 4628 4003
7
SIZES (YNH24 AND MS43a) RANKED
Observed ranked fragment sizes (0 13932 13856 14008 14060 10405 10561 10267 10398 10144 10148 10187 10089 8935 8935 8927 8871 5252 5266 5331 5284 5163 5235 5186 5149 4786 4815 4771 4764 4664 4688 4726 4654 4622 4564 4584 4633 4025 3998 4048 3996 3196 3244 3226 3252 2903 2904 2896 2939 2850 2841 2860 2866 2747 2724 2735 2776 1928 1930 1967
1
Laboratory
FRAGMENT
TABLE
13705 10247 10070 8837 5274 5206 4805 4690 4610 4025 3254 2919 2851 2735
8
13825
9
IN DESCENDING
14087 10560 10157 8942 5240 5173 4772 4668 4625 4034 3260 2942 2859 2766
10
13822 10373 10107 8830 5276 5211 4801 4689 4625 4031 3234 2949 2854 2765 1926
11
13416 10175 9781 8868 5219 5179 4762 4675 4618 3993 3254 2908 2861
13
of variation
13896 10399 10105 8913 5273 5192 4784 4686 4612 4015 3240 2918 2857 2753 1939
BY CRSE
= 0.72
1.58 1.39 1.26 0.58 0.62 0.53 0.43 0.48 0.45 0.46 0.55 0.62 0.31 0.59 0.89
Coefficient of variatian’ (Yo)
RESTRICTED
Between-lub mean (EI)
SAMPLES
Mean coefficient
14911 10943 10648 9310 5311 5210 4803 4698 4622 4040 3214 2938 2838 2744 1941
12
ORDER FOR EACH LABORATORY.
no.
2
0.49
0.28
Deviation (6) fi-omthe 0.29 -0.26 1.27 -0.06 0.17 -0.38 0.47 -0.25 0.39 0.13 0.83 0.56 0.42 -0.04 -0.05 0.46 1.03 0.60 0.48 0.43 -0.13 0.42 0.47 0.75 0.26 0.57 1.06 0.66 0.44 0.54
1
~u.ooracoTy
4
-0.66
-0.81 0.01 -0.42 -0.25 -1.11 -0.83 -0.65 -0.86 -0.46 -0.81 -0.38 -0.73 -0.30 -0.82 1.47
-0.09
-1.18 -1.56 -0.80 -0.16 -0.21 0.11 0.27 0.68 -0.23 -0.24 1.35 0.51 -0.09 0.23
wwan: (%) 6 = (iï? -
3
0.00
0.10 0.36 -0.16 0.03 -0.62 -0.32
0.59
FyM x 100
6
-0.68
-2.54 -2.00 -1.62 -1.19 -0.75 -0.64 -0.50 -0.67 -0.36 0.31 0.24 0.58 0.01 -0.13 -1.01
7
0.24
1.37 1.46 0.36 0.85 -0.03 -0.27 -0.44 -0.09 0.03 -0.24 -0.44 -0.04 0.22 0.66
8
0.51
0.51
9
0.53 0.25 -0.01 0.37 -0.06 -0.37 -0.36 -0.07 -0.29 -0.39 0.18 -1.07 0.12 -0.42 0.64
11
-0.35
-0.06
Mean laboratory
-1.37 -1.55 -0.51 -0.33 0.62 0.36 0.25 0.38 -0.29 -0.46 -0.63 -0.83 -0.06 -0.46
10
3.45 2.16 3.21 0.50 1.02 0.25 0.46 0.23 -0.14 0.56 -0.44 0.34 -0.13 -0.46 0.85
13
-1.60
0.79
dewiation (0) (%)
-7.30 -5.23 -5.36 -4.46 -0.73 -0.35 -0.40 -0.26 -0.23 -0.61 0.79 -0.69 0.68 0.34 -0.13
12
2
3989 3202 2893 2831
2731 1931
3995 3229 2906 2854
2763 1935
5209 4782 4697 4613
2856 2754 1958
4024 3247
2698
4009 3187 2904 2810
ranked jhgmmt sizes (F) 13951 i3943 13914 10278 10430 10378 10098 10066 10198 8800 8914 8858 5241 5285 5243 5122 5171 5081 4773 4710 4796 4640 4699 4587 4622 4548 4582
Observed 13948 10413 10140 8905 5278
4
6
2936
4015
AND MS43a)
2
P)
(YNH24
1
FRAGMENT SIZES LABORATORY
TABLE
14151 10503 10166 8851 5310 5186 4779 4659 4609 4002 3202 2887 2815 2723 1931
7
RANKED
1950
4002 3241 2883 2844
4783 4649
10088 8867 5233
13849
8
2844 2719 1916
3221
4576
5123 4733
10277 10146 8925
9
IN DESCENDING
14092 t0403 10093 8808 5231 5110 4734 4627 4574 4048 3213 2916 2837 2739 1920
10
ORDER
13889 10308 10030 8735 5267 5102 4709 4628 4549 4017 3220 2932 2831 2724 1930
11
3991 3265 2898 2837 2770 1904 Mean coefficient
2858 2772 1953
3223
of variation
5143 4759 4647 4589 4009 3223 2906 2836 2736 1931
13921 10365 10116 8858 5261
Between-lab mean (M,
13
13555 10292 10132 8916 5264 5180 4787 4656 4627
SAMPLES
LABORATORY.
15222 11137 10562 9041 5291 5243 4776 4714 4632
12
FOR EACH
= 0.67
0.42 0.69 0.61 0.50 0.80 0.81
1.13 0.73 0.47 0.66 0.48 0.82 0.67 0.69 0.62
(W
Coemt of variation
RESTRICTED
BY
-0.46
-0.71 -0.67 -1.42
-0.16 -0.63 -0.81 -0.63 -0.45 -0.55 -0.79 -1.12 -0.72 -0.37 -0.75 1.11 0.07 0.91 1.38
0.05 -0.13 0.17 0.00 0.35 1.20 1.02 1.29 0.89
-0.59
-1.03
-0.14
mean: 6 (%) = (fi - F)Ii@ x 100
deviation (0) (070) -0.70 0.64 0.30
Mean laboratory
Deuiation (6) from the -0.19 -0.21 -0.47 0.84 -0.24 0.49 -0.53 0.65 -0.32 0.39 -1.29 0.40 -0.49 -0.30 -1.07 0.15 -0.53 0.15 0.35 0.50 -0.20 0.64 0.00 0.45 -0.64 0.17 -1.00 0.17 -0.23 -0.02
-0.24
-1.33 -0.50 0.08 -0.92 -0.84 -0.43 -0.26 -0.44 0.18 0.64 0.66 0.74 0.46 -0.02
- 1.65
-0.02
-1.01
0.18 -0.57 0.80 -0.29
-0.51 -0.04
0.27 -0.10 0.54
0.52
0.23
-0.29 0.61 0.75
0.05
0.38 0.54 0.39 0.28
0.85 -0.30 0.76
0.07
-1.23 -0.37 0.22 0.56 0.58 0.64 0.52 0.43 0.32 -0.97 0.30 -0.34 -0.04 -0.12 0.55
0.38
0.23 0.55 0.85 1.39 -0.11 0.79 1.04 0.41 0.87 -0.19 0.08 -0.89 0.17 0.43 0.03
-9.34
-2.45
-0.78 -1.33 -1.16
-0.01
-7.45 -4.41 -2.07 -0.56 -1.95 -0.37 -1.44 -0.94
-0.03
2.63 0.70 -0.16 -0.66 -0.05 -0.73 -0.60 -0.19 -0.83 0.45 -1.31 0.28 -0.04 -1.26 1.38
38
2
2.4
2.8
3.2
3.8
4
4.4
4.8
5.2
5.8
MATCH CRITERION (%)
Fig. 3. Interlaboratory comparison of a variable match criterion (kb%) for probe MS43a. ?? , sample A; $, sample B; 0,sample C; A, K562.
100
99 P E
98
R C
97
E N T A G
98 95
E 94 M A T C H
93 92 91 90
2
2.4
2.8
3.2
3.8
4.4
4.8
5.2
5.8
8
MATCH CRITERION (%)
Fig. 4. Interlaboratory comparisons of a variable match criterion (kb%) for probe YNH24. W, sample A; $, sample B; 0,sample C; A, K562; x, experiment 1, sample 12, [ll.
39
If a match guideline is set at 2% then sample 1, band 1 and sample 2, band 1 would be deemed a match provided that each was within the range of * 1% of the mean band weight @b) of the two fragments, i.e. zb = (4564 + 4584)/2 and the match window is therefore xb f (l/lOO) * xb. Of course it would also be a requirement for the second band in both samples to be within the guideline before a match was declared. In this exercise datafiles were produced, (one for YNH24 and the other for MS43a). In each datafile there were up to 22 observations (from 11 laboratories). In Figs. 3 and 4 a match criterion was set between 2-8%. If a comparison of fragments from samples 1 and 2 was within the match criterion then a match was recorded. Sample 1 was then compared with sample 3 and so on to the end of the file. If every sample was compared in this way there was a total of n * (n - 1)/2 comparisons, i.e. up to 231 in this example. This process effectively simulates what would happen if a large number of casework samples were analysed in different laboratories and then compared with each other. The Home Office Forensic Science Service uses a match guideline of 2.8% for the reasons explained by Gil1 et al. [5]. Using this guideline approximately 97.9% of samples probed with YNH24 match between gels and between six different U.K. laboratories [5] al1 using the same protocol. It can be seen in Fig. 4 that use of the European protocol produced results which were comparable to (or perhaps better) than those detailed by Evett et al. [4] and Gil1 et al. [5] because a 2.8% guideline resulted in 799% matches for al1 YNH24 samples tested. In general, use of a 2.8% window with MS43a data was not as defmitive. This was to be expected because the variation in measurement error increased with molecular weight (Tables 1 and 2; Gil1 et al. [6]). This is a useful but simple exercise which laboratories wishing to compare results can follow. It requires only basic computing expertise to write the necessary program. So far only the quality of the match has been considered in this paper (i.e. the numerator of a Bayesian likelihood ratio). Evett et al. [7] have examined the combined effects of the numerator, denominator (i.e. band frequency) and correlation coefficient, by reference to the database of the Metropolitan Police Forensic Science Laboratory (MPFSL). The current MPFSL protocol and the European protocols are very similar; comparisons between the two have not demonstrated differences in fragment size estimations (unpublished results), hence interpretation of the EDNAP data by reference to the MPFSL-derived database and associated statistical parameters is a valid approach and gives a good indication of the power of the current system and its likely success if implemented throughout Europe. Evett et al. [7] conclude that with YNH24 and MS43a using the current EDNAP protocol, between laboratory comparisons of the same DNA sample would result in a likelihood ratio 7 100, with 50% of matches resulting in a likelihood ratio 7 105 (excluding laboratory 12 results). Since completing the analysis an additional laboratory (laboratory 5) has submitted results which have not been included in the current analysis. However, visual inspection of the data shows that their results were wel1 within the match guidelines proposed.
40
A comparison with results of the first EDNAP exercise Results of the first EDNAP exercise were reported in detail by Schneider et al. [ll. When laboratories followed different protocols, up to 10% kb variation of measurements of some fragments were observed. However, within-laboratory variation was much lower and comparable to the results of the (present) experiments reported here. Interlaboratory comparisons have been carried out using the method described in a previous section (An analysis of data using a variable match criterion) for sample 12 (YNH24) from the previous EDNAP experiment [ll. There were 70 sets of samples shared between 9 different laboratories using different protocols. The 2.8% guideline resulted in only 77.9% of matches, a guideline of 5.4% was needed to reach ~97.5% matches in the first EDNAP experiment. Withinlaboratory variation was relatively low, six out of eight laboratories achieved 100% matches using a guideline of 2%. The high interlaboratory variation was a reflection of the variation in the mean molecular sizes of fragments recorded from individual laboratories. Therefore match guidelines needed in the second series of experiments reported here were much lower than those needed in the first series and are comparable to those used in the Forensic Science Service (FSS) of the U.K. at present. They are better applied to lower molecular weight alleles found using YNH24; higher molecular weights of fragments detected using MS43a (> 10 kb) show increased errors of measurement. Conclusion Development of a quality control system This experiment has shown that significant advantages arise from utilisation of a common protocol. It was demonstrated that uniformity could be achieved using a borate buffer system which was free of ethidium bromide. It would be naive to believe that other parameters do not have an effect, but it was encouraging that uniformity was obtained independently of equipment (gel tanks, powerpacks, etc.) used. However, it is possible that buffer temperature variation does affect the mobility of DNA fragments and this is currently under investigation. Experiments to determine the effect of gel length, run-time and make of agarose have demonstrated that these parameters had little effect (FSS, unpublished results); also comparisons between gels run at 4°C and at room temperature have little or no effect (P. Martin, pers. commun.); the constitution of the electrophoretic buffer was the single most important parameter. Once a protocol has been established and agreed, then laboratories need a method to monitor the results which they achieve using an agreed quality control system which incorporates universal ladder markers and genomic controls for sizing purposes. This wil1 enable confident interpretation between laboratories, since each autoradiograph effectively contains a universal control which can be independently checked. It would be necessary for the size of the control to fa11 within designated limits before inclusion into a database. Provided these simple rules are followed there would be no reason why laboratories could not exchange information and combine databases, although the latter would be dependent upon considerations relating to population structure.
41
The achievement of uniformity and quality control methods are currently the subject of active discussion within EDNAP and within the international DNA commission of the International Society for Forensic Haemogenetics. Acknowledgements The authors are indebted to J.E. Lygo, K.L. Faulkner and E.S. Millican of the Forensic Science Service, Centra1 Research and Support Establishment, Aldermaston, U.K. for their significant contribution to the practical and organisational work of this study which ensured that it al1 ran smoothly. Thanks are also due to the following people who contributed to the successful execution of the exercise: P.J. Lincoln (London); P.D. Martin and M. Greenhalgh (Metropolitan Police Forensic Science Laboratory London); E. Valverde (Spain); S. Rand (Munster); B. Olaisen and R. Jonassen (Oslo); E. d’Aloja, M. Dobosz and M. Pescarmona (Rome); A. Kratzer (Zurich); N. Morling (Copenhagen); P. Mangin (Strasbourg). References P.M. Schneider, R. Fimmers, S. Woodroffe, D.J. Werrett, W. Bär, B. Brinkmann, B. Eriksen, S. Jones, A.D. Kloosterman, B. Mevag, V.L. Pascali, C. Rittner, H. Schmitter, J.A. Thomson and P. Gill, Report of a European collaborative exercise comparing DNA typing results using a single lotus VNTR probe. Forensti Sci. Int., 49 (1991) 1- 15. Y. Nakamura, M. Leppert, P. O’Connel, R. Wolff, T. Holm, M. Culver, C. Martin, E. Fujimoto, M. Hoff, E. Kumlin and R. White, Variable number of tandem repeats (VNTR) markers for human gene mapping. Science, 235 (1987) 1616- 1622. 2. Wong, V. Wilson, 1. Patel, S. Povey and A.J. Jeffreys, Characterisation of a panel of highly variable minisatellites cloned from human DNA. Ann. Hum. Genet., 51 (1987) 269 - 288. I.W. Evett, D.J. Werrett, R. Pinchin and P. Gill, Bayesian analysis of single lotus DNA profiles. Proc. Int. Symp. Human Zdentification, 1989, 1990, pp. 77 - 101. P. Gill, I.W. Evett, S. Woodroffe, J.E. Lygo, E. Millican and M. Webster, Databases, quality control and interpretation of DNA profiling in the Home Office Forensic Science Service. Electrw_zhw-eti, 12 (1991) 204 - 209. P. Gill, K. Sullivan and D.J. Werrett, The analysis of hypervariable DNA profiles: problems associated with tbe objective determination of the probability of a match. Hum. Gen&, 85 (1990) 75 - 79. I.W. Evett, J. Scranage and R. Pinchin, Efficient retrieval from DNA databases: based on the second European DNA Profiling Group collaborative exercise. Forensic Sci. Int., 53 (1992) 45-50.
42
Appendices
APPENDIX
1
MOLECULAR
WEIGHTS
OF BANDS OBTAINED
WHEN HYBRIDISED
WITH MS43a
CRSE A, sample A restricted at CRSE; CRSE B, sample B restricted at CRSE; CRSE C, sample C restricted at CRSE; Lab. A, sample A restricted in participating laboratory; Lab. B, sample B restricted in participating laboratory; Lab. C, sample C restricted in participating laboratory; HMwt, High Molecular weight; LMwt, Low Molecular weight. Lab.
1 2 3 4 6 7 8 9 10 11 12 13
CRSE A
CRSE B
Lab. A
Lab. B
HMwt
LMwt
HMwt
LMwt
HMwt
Lifwt
HMwt
LMwt
CRSE C ~ HMwt
10267 10405 10398 10561
5149 5163 5235 5186
10413 10278 10430 10378
5209 5122 5171 5081
10089 10144 10148 10187
4688 4664 4726 4654
10140 10066 10198 10098
4697 4640 4699 4587
8871 8935 8935 8927
8905 8800 8914 8858
10607 10247
5225 5206
10503
5186
10270 10070
4717 4690
5173 5211 5210 5179
10277 10403 10308 11137 10292
5123 5110 5102 5243 5180
10157 10107 10648 9781
4668 4689 4698 4675
4659 4649 4629 4627 4628 4714 4656
9019 8837
10560 10373 10943 10175
10166 10088 10146 10093 10030 10562 10132
8851 8867 8925 8808 8735 9041 8916
APPENDIX
8942 8880 9310 8868
Lab. C LMwt
2
MOLECULAR
WEIGHTS
OF BANDS OBTAINED
WHEN HYBRIDISED
WITH YNH24
CRSE A, sample A restricted at CRSE; CRSE B, sample B restricted at CRSE; CRSE C, sample C restricted at CRSE; Lab. A, sample A restricted in participating Iaboratory; Lab. B, sample B restricted in participating laboratory; Lab. C, sample C restricted in participating laboratory; HMwt, High MolecuIar weight; LMwt, Low Molecular weight. La6
CRSE A
Lab. A
CRSE B
HMwt LMwt
HMwt LMwt
HMwt
LMwt
HMwt
LMwt
HMwt
LMwt
Lab. B
Lab. C
CRSE C
HM&
LMwt
1
4564
2724
4613
2763
4764
1930
4782
1935
3244
2850
3229
2854
2
4534
2735
4582
2731
4786
1928
4713
1931
3226
2841
3202
2831
3
4633
2776
4622
2754
4815
1967
4796
1958
3252
2866
3247
2856
4
4622
2747
4548
2698
4771
4710
3196
2860
3187
2810
6
4607
2762
4576
2761
4756
4744
1924
3245
2875
3226
2847
7
4628
2757
4609
2723
4808
4779
1931
3232
2857
3202
2815
8
4610
2735
4783
1950
3254
2851
3241
2844
4733
1916
3221
2844
4734
1920
3260
2859
3213
2837
9
1958
4805 4576
2719
10
4625
2766
4574
2739
4772
11
4625
2765
4549
2724
4801
1926
4709
1930
3234
2854
3220
2831
12
4622
2744
4632
2772
4803
1941
4776
1953
3214
2838
3223
2858
13
4618
2766
4627
2770
4762
1922
4787
1904
3254
2861
3265
2837
43
APPENDIX MOLECULAR HYBRIDISED
3 WEIGHTS OF BANDS FROM CELL LINE SAMPLE WITH YNH24 AND MS43a
K562 OBTAINED
WHEN
For each laboratory the first line gives resuks for samples placed on the CRSE-restricted half of the gel whereas the second line gives results for samples placed on the ‘in-house’ restricted half of the gel. nd, not determined. Lab.
YNH@
MS@a
Cel1 Lim
Cel1 Lim
HMwt
LMwt
HMwt
LMwt
1
3996 3995
2904 2906
13856 13948
5252 5278
2
3998 3989
2896 2893
13932 13951
5266 5241
3
4048 4024
nd
14008 13943
5331 5285
4
4025 4009
2903 2904
14060 13914
5284 5243
6
4001 4015
2917
nd nd
nd nd
7
4003
2901
14249
5312
4002
2887
14151
5310
8
4025 4002
2919 2883
13705 13849
5274 5233
9
nd
nd
13825
nd
10
4034 4048
2942 2916
14087 14092
5240 5231
11
4031 4017
2949 2932
13822 13889
5276 5267
12
4040
2938
14911 15222
5311
13416 13555
5219 5264
13
3993 3991
2908
5291
a series of samples using its own protocols (the only standardisation was the use of HinfI as the restriction enzyme and use of a common ladder marker). It was demonstrated that the sizes of fragments determined by different laboratories were within a match window of 10%. Although it was possible to compare directly results between laboratories using this wide window, interpretation and allocation of statistical significante was more difficult when such large differences were obtained between laboratories. The second series of experiments described in this paper was carried out in order to determine whether comparable results could be obtained if different laboratories used the same protocol. Clearly, it was not possible to standardise completely because different equipment was in use. However, it was possible to standardise on electrophoretic buffer and make of agarose. Also each laboratory was supplied with a full protocol and samples of DNA; equipment was not standardised. A comparison of the within- and between-laboratory measurement error has been evaluated. Materials and Methods
DNA was bulk-extracted by one laboratory (Centra1 Research & Support Establishment CRSE) from three different blood samples. Half was then restricted by this laboratory with 30 x excess Hinff before distribution and the other half was supplied to laboratories unrestricted so that the effect of different restriction methods could be compared. In addition, HinfI restricted K562 (genomic control; Promega) and Amersham lambda ladder was supplied to each laboratory. Each laboratory was supplied with a full protocol; ICN Seakem GTG agarose was distributed from the same bottle. The agarose (2.2 g) was added to 315 ml Tris-borate EDTA (TBE) (10 x TBE = 1.3 M Tris, 0.75 M boric acid, 25 mM ethylenediaminetetraacetic acid (EDTA); pH 8.8) to make a 20 x 25 cm gel. Gels were electrophoresed in TBE buffer at 55 V until the 2 kb marker of an ethidium bromide-saturated X HindIII marker had travelled 16 cm along the gel (otherwise the electrophoretic system was ethidium bromide-free). Southern blots were probed with YNH24 [2] and MS43a [3]; fragments were sized at CRSE using Biotrac (Foster and Freeman Ltd). The experimental design is illustrated in Fig. 1. Each laboratory followed the protocol supplied. Both unrestricted and restricted DNA was supplied by the coordinating laboratory. Each participating laboratory was asked to restrict DNA supplied using their usual method. Results and Discussion Differences between samples restricted in-heuse compared to those restricted in the participating laboratmy
Minor differences were observed between samples restricted by CRSE compared to samples restricted in the participating laboratory (Appendices 1 - 3).
4.2
3.5
2.7
Fig. 1. Photograph of an autoradiograph showing the experimental design. The left hand side of the autoradiograph consisted of samples restricted by CRSE whereas the right hand side consisted of samples restricted by the participating laboratory. h, ladder marker; g, genomic control (not used in this exercise); a, sample a; b, sample b; c, sample c; k, K56.2.
32
Band positions on each half of the gel for each laboratory were compared using a Wilcoxon sign rank test i.e. laboratory 1 CRSE restricted samples A, B, C, K562 (YNH24 and MS43a) were compared with laboratory 1 restricted samples A, B, C, K562 (YNH24 and MS43a). Remaining laboratories were tested separately. Results were usually significant: the CRSE samples were restricted at the same time (and in the same way) on both halves of the gel, but stil1 gave a significantly different result. The differente was probably systematic. The reason for the effect was almost certainly because of the linear interpolation method for sizing bands between control ladder markers. The schematic diagram in Fig. 2 shows a greatly exaggerated distortion (purely for illustrative purposes) in both halves of the electrophoretic plate. In this example, the control sample is displaced in opposite directions in the right hand side compared to the left hand side of the plate because of curvilinear electrophoretic distortion; hence the systematic band shifts are correlated. In the current experiment (Fig. l), the differences between fragment sizes from each half of the gel were low (< 0.5% kb); it is very unlikely that different restriction buffers or different manufacturer’s í%~fI restriction enzyme contribute to the effect.
-
-m
_---
-_
___d
--
_ _
&
-
-_
-.
-4
r--
s---w_
-4
--_
--___-4 _M---_ ti--@
---___d
_-----. -.
__-
&#
/-
F
-w-
‘-
Fig. 2. A schematic diagram showing reasons for correlated shift due to electrophoretic phenomenon. The linear interpolation method does not take account of curvilinear distortion in the gel, hence in the above example (greatly exaggerated) both bands are shifted towards the high molecular linear interpolation. weight region in the right hand side of the gel. (- - - -), distortion; (-),
33
Calculation of okviation frorn the interlaboratory mean Laboratory 12 showed significant deviation from the mean, particularly for high molecular weight fragments >9.3 kb. This was probably because electrophoresis was carried out on a particularly hot day when buffer temperature could have reached >3O”C. Results from this laboratory were excluded when calculating interlaboratory means and coefficients of variation. Even so, results obtained in laboratory 12 from the lower molecular weight fragments detected by YNH24 were comparable to those from other laboratories. In Tables 1 and 2, fragments from both YNH24 and MS43a are given. They are arranged in descending size order for each laboratory. Between-laboratory means (M> and coefficients of variation were calculated. For each fragment, deviation from the mean (ìï@ was estimated and the data were summarised to show average laboratory deviation. Results from each half of the autoradiograph were compared (i.e. one half contained samples restricted by CRSE, Table 1 and the other half contained samples restricted by the participating laboratory, Table 2). The between-laboratory coefficient of variation was greatest for high molecular weight fragments (lO- 14 kb); an overall average of 0.72% kb was recorded for the whole range in Table 1. The mean laboratory deviation was no greater than 0.8% kb; this compared with an approximate 5.15% kb (YNH24) variation observed in the first series of EDNAP experiments for one sample, number 12 [ll. There were no obvious differences seen when testing samples restricted by the participating laboratories as compared to samples restricted by CRSE (between-laboratory mean values (IM>were very similar and no greater than 0.2% kb different from each other). Some laboratories showed predominantly negative deviation from the mean (e.g. laboratories 3 and 12; Table 1) whereas others were positive (laboratory 2, Table 1); laboratories 1 and 13 showed positive deviation in Table 1 and negative deviation in the second half of the autoradiograph (Table 2); this was a reflection of correlation of fragment deviation dependent upon unpredictable minor electrophoretic differences between two halves of a gel. An analysis of data wing a variable match criterion Evett et al. [4] introduced a method of analysis which enabled a large number of comparisons to be carried out from a relatively smal1 database. The principle of the test involves comparison of each pair of band measurements with every other pair in the database referring to a match guideline. Part of the datafile of sample A (YNH24) is as follows: Identifier (laboratory)
Band Weights (bp) 1 2
1
4564 4584 4633
2 3
2724 2735 2776
1
2
no.
3
4
4607 4001 3245 2917 2875 2762
4756
6
2901 2857 2757 1958
14249 10607 10270 9019 5312 5225 4808 4717 4628 4003
7
SIZES (YNH24 AND MS43a) RANKED
Observed ranked fragment sizes (0 13932 13856 14008 14060 10405 10561 10267 10398 10144 10148 10187 10089 8935 8935 8927 8871 5252 5266 5331 5284 5163 5235 5186 5149 4786 4815 4771 4764 4664 4688 4726 4654 4622 4564 4584 4633 4025 3998 4048 3996 3196 3244 3226 3252 2903 2904 2896 2939 2850 2841 2860 2866 2747 2724 2735 2776 1928 1930 1967
1
Laboratory
FRAGMENT
TABLE
13705 10247 10070 8837 5274 5206 4805 4690 4610 4025 3254 2919 2851 2735
8
13825
9
IN DESCENDING
14087 10560 10157 8942 5240 5173 4772 4668 4625 4034 3260 2942 2859 2766
10
13822 10373 10107 8830 5276 5211 4801 4689 4625 4031 3234 2949 2854 2765 1926
11
13416 10175 9781 8868 5219 5179 4762 4675 4618 3993 3254 2908 2861
13
of variation
13896 10399 10105 8913 5273 5192 4784 4686 4612 4015 3240 2918 2857 2753 1939
BY CRSE
= 0.72
1.58 1.39 1.26 0.58 0.62 0.53 0.43 0.48 0.45 0.46 0.55 0.62 0.31 0.59 0.89
Coefficient of variatian’ (Yo)
RESTRICTED
Between-lub mean (EI)
SAMPLES
Mean coefficient
14911 10943 10648 9310 5311 5210 4803 4698 4622 4040 3214 2938 2838 2744 1941
12
ORDER FOR EACH LABORATORY.
no.
2
0.49
0.28
Deviation (6) fi-omthe 0.29 -0.26 1.27 -0.06 0.17 -0.38 0.47 -0.25 0.39 0.13 0.83 0.56 0.42 -0.04 -0.05 0.46 1.03 0.60 0.48 0.43 -0.13 0.42 0.47 0.75 0.26 0.57 1.06 0.66 0.44 0.54
1
~u.ooracoTy
4
-0.66
-0.81 0.01 -0.42 -0.25 -1.11 -0.83 -0.65 -0.86 -0.46 -0.81 -0.38 -0.73 -0.30 -0.82 1.47
-0.09
-1.18 -1.56 -0.80 -0.16 -0.21 0.11 0.27 0.68 -0.23 -0.24 1.35 0.51 -0.09 0.23
wwan: (%) 6 = (iï? -
3
0.00
0.10 0.36 -0.16 0.03 -0.62 -0.32
0.59
FyM x 100
6
-0.68
-2.54 -2.00 -1.62 -1.19 -0.75 -0.64 -0.50 -0.67 -0.36 0.31 0.24 0.58 0.01 -0.13 -1.01
7
0.24
1.37 1.46 0.36 0.85 -0.03 -0.27 -0.44 -0.09 0.03 -0.24 -0.44 -0.04 0.22 0.66
8
0.51
0.51
9
0.53 0.25 -0.01 0.37 -0.06 -0.37 -0.36 -0.07 -0.29 -0.39 0.18 -1.07 0.12 -0.42 0.64
11
-0.35
-0.06
Mean laboratory
-1.37 -1.55 -0.51 -0.33 0.62 0.36 0.25 0.38 -0.29 -0.46 -0.63 -0.83 -0.06 -0.46
10
3.45 2.16 3.21 0.50 1.02 0.25 0.46 0.23 -0.14 0.56 -0.44 0.34 -0.13 -0.46 0.85
13
-1.60
0.79
dewiation (0) (%)
-7.30 -5.23 -5.36 -4.46 -0.73 -0.35 -0.40 -0.26 -0.23 -0.61 0.79 -0.69 0.68 0.34 -0.13
12
2
3989 3202 2893 2831
2731 1931
3995 3229 2906 2854
2763 1935
5209 4782 4697 4613
2856 2754 1958
4024 3247
2698
4009 3187 2904 2810
ranked jhgmmt sizes (F) 13951 i3943 13914 10278 10430 10378 10098 10066 10198 8800 8914 8858 5241 5285 5243 5122 5171 5081 4773 4710 4796 4640 4699 4587 4622 4548 4582
Observed 13948 10413 10140 8905 5278
4
6
2936
4015
AND MS43a)
2
P)
(YNH24
1
FRAGMENT SIZES LABORATORY
TABLE
14151 10503 10166 8851 5310 5186 4779 4659 4609 4002 3202 2887 2815 2723 1931
7
RANKED
1950
4002 3241 2883 2844
4783 4649
10088 8867 5233
13849
8
2844 2719 1916
3221
4576
5123 4733
10277 10146 8925
9
IN DESCENDING
14092 t0403 10093 8808 5231 5110 4734 4627 4574 4048 3213 2916 2837 2739 1920
10
ORDER
13889 10308 10030 8735 5267 5102 4709 4628 4549 4017 3220 2932 2831 2724 1930
11
3991 3265 2898 2837 2770 1904 Mean coefficient
2858 2772 1953
3223
of variation
5143 4759 4647 4589 4009 3223 2906 2836 2736 1931
13921 10365 10116 8858 5261
Between-lab mean (M,
13
13555 10292 10132 8916 5264 5180 4787 4656 4627
SAMPLES
LABORATORY.
15222 11137 10562 9041 5291 5243 4776 4714 4632
12
FOR EACH
= 0.67
0.42 0.69 0.61 0.50 0.80 0.81
1.13 0.73 0.47 0.66 0.48 0.82 0.67 0.69 0.62
(W
Coemt of variation
RESTRICTED
BY
-0.46
-0.71 -0.67 -1.42
-0.16 -0.63 -0.81 -0.63 -0.45 -0.55 -0.79 -1.12 -0.72 -0.37 -0.75 1.11 0.07 0.91 1.38
0.05 -0.13 0.17 0.00 0.35 1.20 1.02 1.29 0.89
-0.59
-1.03
-0.14
mean: 6 (%) = (fi - F)Ii@ x 100
deviation (0) (070) -0.70 0.64 0.30
Mean laboratory
Deuiation (6) from the -0.19 -0.21 -0.47 0.84 -0.24 0.49 -0.53 0.65 -0.32 0.39 -1.29 0.40 -0.49 -0.30 -1.07 0.15 -0.53 0.15 0.35 0.50 -0.20 0.64 0.00 0.45 -0.64 0.17 -1.00 0.17 -0.23 -0.02
-0.24
-1.33 -0.50 0.08 -0.92 -0.84 -0.43 -0.26 -0.44 0.18 0.64 0.66 0.74 0.46 -0.02
- 1.65
-0.02
-1.01
0.18 -0.57 0.80 -0.29
-0.51 -0.04
0.27 -0.10 0.54
0.52
0.23
-0.29 0.61 0.75
0.05
0.38 0.54 0.39 0.28
0.85 -0.30 0.76
0.07
-1.23 -0.37 0.22 0.56 0.58 0.64 0.52 0.43 0.32 -0.97 0.30 -0.34 -0.04 -0.12 0.55
0.38
0.23 0.55 0.85 1.39 -0.11 0.79 1.04 0.41 0.87 -0.19 0.08 -0.89 0.17 0.43 0.03
-9.34
-2.45
-0.78 -1.33 -1.16
-0.01
-7.45 -4.41 -2.07 -0.56 -1.95 -0.37 -1.44 -0.94
-0.03
2.63 0.70 -0.16 -0.66 -0.05 -0.73 -0.60 -0.19 -0.83 0.45 -1.31 0.28 -0.04 -1.26 1.38
38
2
2.4
2.8
3.2
3.8
4
4.4
4.8
5.2
5.8
MATCH CRITERION (%)
Fig. 3. Interlaboratory comparison of a variable match criterion (kb%) for probe MS43a. ?? , sample A; $, sample B; 0,sample C; A, K562.
100
99 P E
98
R C
97
E N T A G
98 95
E 94 M A T C H
93 92 91 90
2
2.4
2.8
3.2
3.8
4.4
4.8
5.2
5.8
8
MATCH CRITERION (%)
Fig. 4. Interlaboratory comparisons of a variable match criterion (kb%) for probe YNH24. W, sample A; $, sample B; 0,sample C; A, K562; x, experiment 1, sample 12, [ll.
39
If a match guideline is set at 2% then sample 1, band 1 and sample 2, band 1 would be deemed a match provided that each was within the range of * 1% of the mean band weight @b) of the two fragments, i.e. zb = (4564 + 4584)/2 and the match window is therefore xb f (l/lOO) * xb. Of course it would also be a requirement for the second band in both samples to be within the guideline before a match was declared. In this exercise datafiles were produced, (one for YNH24 and the other for MS43a). In each datafile there were up to 22 observations (from 11 laboratories). In Figs. 3 and 4 a match criterion was set between 2-8%. If a comparison of fragments from samples 1 and 2 was within the match criterion then a match was recorded. Sample 1 was then compared with sample 3 and so on to the end of the file. If every sample was compared in this way there was a total of n * (n - 1)/2 comparisons, i.e. up to 231 in this example. This process effectively simulates what would happen if a large number of casework samples were analysed in different laboratories and then compared with each other. The Home Office Forensic Science Service uses a match guideline of 2.8% for the reasons explained by Gil1 et al. [5]. Using this guideline approximately 97.9% of samples probed with YNH24 match between gels and between six different U.K. laboratories [5] al1 using the same protocol. It can be seen in Fig. 4 that use of the European protocol produced results which were comparable to (or perhaps better) than those detailed by Evett et al. [4] and Gil1 et al. [5] because a 2.8% guideline resulted in 799% matches for al1 YNH24 samples tested. In general, use of a 2.8% window with MS43a data was not as defmitive. This was to be expected because the variation in measurement error increased with molecular weight (Tables 1 and 2; Gil1 et al. [6]). This is a useful but simple exercise which laboratories wishing to compare results can follow. It requires only basic computing expertise to write the necessary program. So far only the quality of the match has been considered in this paper (i.e. the numerator of a Bayesian likelihood ratio). Evett et al. [7] have examined the combined effects of the numerator, denominator (i.e. band frequency) and correlation coefficient, by reference to the database of the Metropolitan Police Forensic Science Laboratory (MPFSL). The current MPFSL protocol and the European protocols are very similar; comparisons between the two have not demonstrated differences in fragment size estimations (unpublished results), hence interpretation of the EDNAP data by reference to the MPFSL-derived database and associated statistical parameters is a valid approach and gives a good indication of the power of the current system and its likely success if implemented throughout Europe. Evett et al. [7] conclude that with YNH24 and MS43a using the current EDNAP protocol, between laboratory comparisons of the same DNA sample would result in a likelihood ratio 7 100, with 50% of matches resulting in a likelihood ratio 7 105 (excluding laboratory 12 results). Since completing the analysis an additional laboratory (laboratory 5) has submitted results which have not been included in the current analysis. However, visual inspection of the data shows that their results were wel1 within the match guidelines proposed.
40
A comparison with results of the first EDNAP exercise Results of the first EDNAP exercise were reported in detail by Schneider et al. [ll. When laboratories followed different protocols, up to 10% kb variation of measurements of some fragments were observed. However, within-laboratory variation was much lower and comparable to the results of the (present) experiments reported here. Interlaboratory comparisons have been carried out using the method described in a previous section (An analysis of data using a variable match criterion) for sample 12 (YNH24) from the previous EDNAP experiment [ll. There were 70 sets of samples shared between 9 different laboratories using different protocols. The 2.8% guideline resulted in only 77.9% of matches, a guideline of 5.4% was needed to reach ~97.5% matches in the first EDNAP experiment. Withinlaboratory variation was relatively low, six out of eight laboratories achieved 100% matches using a guideline of 2%. The high interlaboratory variation was a reflection of the variation in the mean molecular sizes of fragments recorded from individual laboratories. Therefore match guidelines needed in the second series of experiments reported here were much lower than those needed in the first series and are comparable to those used in the Forensic Science Service (FSS) of the U.K. at present. They are better applied to lower molecular weight alleles found using YNH24; higher molecular weights of fragments detected using MS43a (> 10 kb) show increased errors of measurement. Conclusion Development of a quality control system This experiment has shown that significant advantages arise from utilisation of a common protocol. It was demonstrated that uniformity could be achieved using a borate buffer system which was free of ethidium bromide. It would be naive to believe that other parameters do not have an effect, but it was encouraging that uniformity was obtained independently of equipment (gel tanks, powerpacks, etc.) used. However, it is possible that buffer temperature variation does affect the mobility of DNA fragments and this is currently under investigation. Experiments to determine the effect of gel length, run-time and make of agarose have demonstrated that these parameters had little effect (FSS, unpublished results); also comparisons between gels run at 4°C and at room temperature have little or no effect (P. Martin, pers. commun.); the constitution of the electrophoretic buffer was the single most important parameter. Once a protocol has been established and agreed, then laboratories need a method to monitor the results which they achieve using an agreed quality control system which incorporates universal ladder markers and genomic controls for sizing purposes. This wil1 enable confident interpretation between laboratories, since each autoradiograph effectively contains a universal control which can be independently checked. It would be necessary for the size of the control to fa11 within designated limits before inclusion into a database. Provided these simple rules are followed there would be no reason why laboratories could not exchange information and combine databases, although the latter would be dependent upon considerations relating to population structure.
41
The achievement of uniformity and quality control methods are currently the subject of active discussion within EDNAP and within the international DNA commission of the International Society for Forensic Haemogenetics. Acknowledgements The authors are indebted to J.E. Lygo, K.L. Faulkner and E.S. Millican of the Forensic Science Service, Centra1 Research and Support Establishment, Aldermaston, U.K. for their significant contribution to the practical and organisational work of this study which ensured that it al1 ran smoothly. Thanks are also due to the following people who contributed to the successful execution of the exercise: P.J. Lincoln (London); P.D. Martin and M. Greenhalgh (Metropolitan Police Forensic Science Laboratory London); E. Valverde (Spain); S. Rand (Munster); B. Olaisen and R. Jonassen (Oslo); E. d’Aloja, M. Dobosz and M. Pescarmona (Rome); A. Kratzer (Zurich); N. Morling (Copenhagen); P. Mangin (Strasbourg). References P.M. Schneider, R. Fimmers, S. Woodroffe, D.J. Werrett, W. Bär, B. Brinkmann, B. Eriksen, S. Jones, A.D. Kloosterman, B. Mevag, V.L. Pascali, C. Rittner, H. Schmitter, J.A. Thomson and P. Gill, Report of a European collaborative exercise comparing DNA typing results using a single lotus VNTR probe. Forensti Sci. Int., 49 (1991) 1- 15. Y. Nakamura, M. Leppert, P. O’Connel, R. Wolff, T. Holm, M. Culver, C. Martin, E. Fujimoto, M. Hoff, E. Kumlin and R. White, Variable number of tandem repeats (VNTR) markers for human gene mapping. Science, 235 (1987) 1616- 1622. 2. Wong, V. Wilson, 1. Patel, S. Povey and A.J. Jeffreys, Characterisation of a panel of highly variable minisatellites cloned from human DNA. Ann. Hum. Genet., 51 (1987) 269 - 288. I.W. Evett, D.J. Werrett, R. Pinchin and P. Gill, Bayesian analysis of single lotus DNA profiles. Proc. Int. Symp. Human Zdentification, 1989, 1990, pp. 77 - 101. P. Gill, I.W. Evett, S. Woodroffe, J.E. Lygo, E. Millican and M. Webster, Databases, quality control and interpretation of DNA profiling in the Home Office Forensic Science Service. Electrw_zhw-eti, 12 (1991) 204 - 209. P. Gill, K. Sullivan and D.J. Werrett, The analysis of hypervariable DNA profiles: problems associated with tbe objective determination of the probability of a match. Hum. Gen&, 85 (1990) 75 - 79. I.W. Evett, J. Scranage and R. Pinchin, Efficient retrieval from DNA databases: based on the second European DNA Profiling Group collaborative exercise. Forensic Sci. Int., 53 (1992) 45-50.
42
Appendices
APPENDIX
1
MOLECULAR
WEIGHTS
OF BANDS OBTAINED
WHEN HYBRIDISED
WITH MS43a
CRSE A, sample A restricted at CRSE; CRSE B, sample B restricted at CRSE; CRSE C, sample C restricted at CRSE; Lab. A, sample A restricted in participating laboratory; Lab. B, sample B restricted in participating laboratory; Lab. C, sample C restricted in participating laboratory; HMwt, High Molecular weight; LMwt, Low Molecular weight. Lab.
1 2 3 4 6 7 8 9 10 11 12 13
CRSE A
CRSE B
Lab. A
Lab. B
HMwt
LMwt
HMwt
LMwt
HMwt
Lifwt
HMwt
LMwt
CRSE C ~ HMwt
10267 10405 10398 10561
5149 5163 5235 5186
10413 10278 10430 10378
5209 5122 5171 5081
10089 10144 10148 10187
4688 4664 4726 4654
10140 10066 10198 10098
4697 4640 4699 4587
8871 8935 8935 8927
8905 8800 8914 8858
10607 10247
5225 5206
10503
5186
10270 10070
4717 4690
5173 5211 5210 5179
10277 10403 10308 11137 10292
5123 5110 5102 5243 5180
10157 10107 10648 9781
4668 4689 4698 4675
4659 4649 4629 4627 4628 4714 4656
9019 8837
10560 10373 10943 10175
10166 10088 10146 10093 10030 10562 10132
8851 8867 8925 8808 8735 9041 8916
APPENDIX
8942 8880 9310 8868
Lab. C LMwt
2
MOLECULAR
WEIGHTS
OF BANDS OBTAINED
WHEN HYBRIDISED
WITH YNH24
CRSE A, sample A restricted at CRSE; CRSE B, sample B restricted at CRSE; CRSE C, sample C restricted at CRSE; Lab. A, sample A restricted in participating Iaboratory; Lab. B, sample B restricted in participating laboratory; Lab. C, sample C restricted in participating laboratory; HMwt, High MolecuIar weight; LMwt, Low Molecular weight. La6
CRSE A
Lab. A
CRSE B
HMwt LMwt
HMwt LMwt
HMwt
LMwt
HMwt
LMwt
HMwt
LMwt
Lab. B
Lab. C
CRSE C
HM&
LMwt
1
4564
2724
4613
2763
4764
1930
4782
1935
3244
2850
3229
2854
2
4534
2735
4582
2731
4786
1928
4713
1931
3226
2841
3202
2831
3
4633
2776
4622
2754
4815
1967
4796
1958
3252
2866
3247
2856
4
4622
2747
4548
2698
4771
4710
3196
2860
3187
2810
6
4607
2762
4576
2761
4756
4744
1924
3245
2875
3226
2847
7
4628
2757
4609
2723
4808
4779
1931
3232
2857
3202
2815
8
4610
2735
4783
1950
3254
2851
3241
2844
4733
1916
3221
2844
4734
1920
3260
2859
3213
2837
9
1958
4805 4576
2719
10
4625
2766
4574
2739
4772
11
4625
2765
4549
2724
4801
1926
4709
1930
3234
2854
3220
2831
12
4622
2744
4632
2772
4803
1941
4776
1953
3214
2838
3223
2858
13
4618
2766
4627
2770
4762
1922
4787
1904
3254
2861
3265
2837
43
APPENDIX MOLECULAR HYBRIDISED
3 WEIGHTS OF BANDS FROM CELL LINE SAMPLE WITH YNH24 AND MS43a
K562 OBTAINED
WHEN
For each laboratory the first line gives resuks for samples placed on the CRSE-restricted half of the gel whereas the second line gives results for samples placed on the ‘in-house’ restricted half of the gel. nd, not determined. Lab.
YNH@
MS@a
Cel1 Lim
Cel1 Lim
HMwt
LMwt
HMwt
LMwt
1
3996 3995
2904 2906
13856 13948
5252 5278
2
3998 3989
2896 2893
13932 13951
5266 5241
3
4048 4024
nd
14008 13943
5331 5285
4
4025 4009
2903 2904
14060 13914
5284 5243
6
4001 4015
2917
nd nd
nd nd
7
4003
2901
14249
5312
4002
2887
14151
5310
8
4025 4002
2919 2883
13705 13849
5274 5233
9
nd
nd
13825
nd
10
4034 4048
2942 2916
14087 14092
5240 5231
11
4031 4017
2949 2932
13822 13889
5276 5267
12
4040
2938
14911 15222
5311
13416 13555
5219 5264
13
3993 3991
2908
5291