Method for determining crystallization parameters and apparatus for use with the same
The present disclosure provides a method to allow a user to pre-screen numerous crystallization conditions in the crystallization space to identify those conditions with the highest probability of yielding crystals and high quality diffracting crystals. In one embodiment, the dilute solution thermodynamic virial coefficient, termed B, is used to aid in the determination crystallization conditions that increase the probability of producing crystals for the crystallant of interest. The present disclosure also provide methods for predicting solution conditions that generate beneficial solubility and/or stability conditions for a polypeptide of interest using the B parameter. Devices for use in the described methods are also described.
1. A method for generating a predicted positive crystallization condition for a crystallant of interest, the method comprising the steps of:
(a) conducting an initial crystallization screen on the crystallant of interest, the initial crystallization screen comprising a plurality of crystallization conditions, each of the crystallization conditions having one or more crystallization condition variables;
(b) determining a crystallization outcome for at least one of the crystallization conditions of the initial crystallization screen;
(c) determining for at least one of the crystallization conditions of the initial crystallization screen a dilute solution thermodynamic virial coefficient (B) parameter;
(d) inputting a data set into a predictive crystallization function, the data set including information selected from the group consisting of: (i) one or more of the crystallization condition variables for at least one of the crystallization condition of the initial crystallization screen, (ii) the crystallization outcome for at least one of the crystallization condition of the initial crystallization screen, (iii) the B parameter for at least one of the crystallization condition of the initial crystallization screen, and (iv) any combination of the foregoing;
(e) training the predictive crystallization function using the data set where the training is continued until the predictive crystallization function reaches a suitable convergence; and
(f) generating the predicted positive crystallization condition for the crystallant of interest using the converged values of crystallization variables in the predictive crystallization function.
2. The method of claim 1 where the predictive positive crystallization condition is not contained in the initial crystallization screen.
3. The method of claim 1 where the crystallization condition variables define a total crystallization space for the crystallant of interest and the crystallization conditions sampling less than 5% of the total crystallization space.
4. The method of claim 1 where the crystallant of interest is selected from the group consisting of a polypeptide, an organic molecule, an organometallic molecule, an inorganic molecule, a nanocrystal, and a virus.
5. The method of claim 4 where the polypeptide is an aqueous polypeptide or a membrane polypeptide.
6. The method of claim 1 where the crystallization condition variables are selected from the group consisting of purity of the crystallant, dilution of the crystallant, homogeneity of the crystallant, stability of the crystallant, flexibility of the crystallant, precipitating agent, concentration of precipitating agent, buffer, concentration of buffer, pH, temperature, light, magnetism, gravity, atmosphere identity, atmospheric pressure, divalent anion, organic moment, additives to aid in crystallization and a combination of any of the foregoing.
7. The method of claim 6 where the crystallization condition variables are further selected from the group consisting of type of lipid present, detergent, concentration of detergent, polar additives, apolar additives, amphipathic additives and a combination of any of the foregoing.
8. The method of claim 3 where the initial crystallization screen comprises less than 4%, less than 3%, less than 2%, less than 1%, less than 0.5% or less than 0.1% of the total crystallization space.
9. The method of claim 1 where the initial crystallization screen is generated using an incomplete factorial screen or a random screen.
10. The method of claim 1 where the predictive crystallization function is a neural network.
11. The method of claim 10 where the neural network is trained through back propagation.
12. The method of claim 1 where the predictive crystallization function is selected from the group consisting of a Chernov algorithm, a Bayesian net, a Bayesian classification, a Bayesian decomposition, a Mahalanobis distance, a Gram-Schmidt algorithm and a cluster analysis.
13. The method of claim 1 further comprising generating crystals of the crystallant of interest using the predicted positive crystallization condition.
14. The method of claim 1 where the B parameter is determined using self-interacting chromatography or static light scattering.
15. A method for generating a predicted positive crystallization condition for a crystallant of interest, the method comprising the steps of:
(a) conducting an initial crystallization screen, the initial crystallization screen comprising a plurality of crystallization conditions, each of the crystallization conditions having one or more crystallization condition variables;
(b) determining a crystallization outcome for at least one of the crystallization conditions of the initial crystallization screen;
(c) inputting a data set into a predictive crystallization function, the data set including information selected from the group consisting of: (i) one or more of the crystallization condition variables for at least one of the crystallization condition of the initial crystallization screen, (ii) the crystallization outcome for at least one of the crystallization condition of the initial crystallization screen, and (iii) any combination of the foregoing;
(d) training the predictive crystallization function using the data set where the training is continued until the predictive crystallization function reaches a suitable convergence;
(e) generating the predicted positive crystallization condition for the crystallant of interest using the converged values of crystallization variables in the trained predictive crystallization function; and
(f) validating the predicted positive crystallization condition using a dilute solution thermodynamic virial coefficient (B) parameter.
16. The method of claim 15 where the validating comprises determining the B parameter for the predicted positive crystallization condition and determining whether the B parameter is within a range that falls inside a crystallization slot.
17. The method of claim 16 where the predicted positive crystallization condition is validated if the B parameter is within a range that falls inside the crystallization slot.
18. The method of claim 16 where the predicted crystallization condition is not validated if the B parameter is not within a range that falls inside the crystallization slot.
19. The method of claim 15 where the B parameter is determined using self-interacting chromatography or static light scattering.
20. The method of claim 15 further comprising generating crystals of the crystallant of interest using the predicted positive crystallization condition that has been validated.
21. The method of claim 15 where the crystallant of interest is selected from the group consisting of a polypeptide, an organic molecule, an organometallic molecule, an inorganic molecule, a nanocrystal, and a virus.