Table 1.
Simulation test on bimodal deuterated distributions.
| Distance between two deuterated numbers in a bimodal distribution | ||||||||
|---|---|---|---|---|---|---|---|---|
| Weight | 2 | 3 | 5 | 8 | ||||
| Initial | Expanded | Initial | Expanded | Initial | Expanded | Initial | Expanded | |
| (a) For peptides of length 10 | ||||||||
| 0.5 | 49% | 100% | 80% | 100% | 100% | 100% | 100% | 100% |
| 0.6 | 68% | 100% | 83% | 100% | 98% | 100% | 100% | 100% |
| 0.7 | 37% | 100% | 86% | 100% | 92% | 100% | 100% | 100% |
| 0.8 | 24% | 100% | 58% | 100% | 91% | 100% | 100% | 100% |
| 0.9 | 18% | 100% | 42% | 98% | 92% | 100% | 100% | 100% |
| (b) For peptides of length 15 | ||||||||
| 0.5 | 54% | 100% | 79% | 100% | 100% | 100% | 100% | 100% |
| 0.6 | 45% | 100% | 85% | 100% | 100% | 100% | 100% | 100% |
| 0.7 | 23% | 100% | 71% | 100% | 100% | 100% | 100% | 100% |
| 0.8 | 18% | 100% | 53% | 100% | 100% | 100% | 100% | 100% |
| 0.9 | 19% | 100% | 43% | 100% | 89% | 94% | 100% | 100% |
| (c) For peptides of length 20 | ||||||||
| 0.5 | 48% | 100% | 75% | 100% | 100% | 100% | 100% | 100% |
| 0.6 | 29% | 100% | 74% | 100% | 100% | 100% | 100% | 100% |
| 0.7 | 21% | 100% | 54% | 100% | 99% | 99% | 100% | 100% |
| 0.8 | 18% | 100% | 45% | 100% | 99% | 100% | 100% | 100% |
| 0.9 | 18% | 100% | 42% | 100% | 83% | 97% | 100% | 100% |
| (d) For peptides of length 30 | ||||||||
| 0.5 | 14% | 100% | 67% | 100% | 99% | 100% | 100% | 100% |
| 0.6 | 21% | 100% | 56% | 100% | 99% | 100% | 100% | 100% |
| 0.7 | 19% | 100% | 47% | 100% | 97% | 98% | 100% | 100% |
| 0.8 | 17% | 100% | 43% | 100% | 87% | 94% | 100% | 100% |
| 0.9 | 24% | 100% | 42% | 100% | 75% | 92% | 100% | 100% |
With varying distances and weights between two distributions in bimodal analysis, deMix performance is shown. Each cell shows how deMix correctly determined two deuterated forms for bimodal distributions simulated under a specific condition. For example, in b), peptides of length 15 were randomly selected from Swiss-Prot human database, and then two deuterium numbers and their respective weights were randomly generated. Based on the simulated values, bimodal deuterated distributions were generated (10,000 different distributions for each cell). When the distance between two deuterium numbers was close, deMix often reported tdA plus 1 for a small true value tdA while tdB minus 1 for a large true value tdB and failed to determine the exact numbers (initial column). Based on the observation, deMix in bimodal analysis takes into account 4 combinations using dA − 1 and dB + 1 in addition to dA and dB from initial prediction. The adaptation led to outstanding performance for all cases (expanded column). In ‘weight’ row, the weight (w) of more abundant form is represented (the weight of the other form is 1-w).