dissenting.
Absent the procedural safeguards provided to the non-moving party at the summary judgment stage, I might very well agree with the majority that the PTO demonstrated the obviousness of these claims. When evaluating a case on appeal from summary judgment, however, “[t]he evidence of the non-movant is to be believed, and all justifiable inferences are to be drawn in his favor.” Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 255, 106-S.Ct. 2505, 91 L.Ed.2d 202 (1986). NantKwest submitted evidence and accompanying expert testimony showing that an ordinarily skilled artisan at the time of the invention would not have had a reasonable expectation of success in combining Santoli’s in vivo results for TALL-104 cells with Gong’s in vitro experiments for NK-92 *875cells. Drawing all reasonable inferences in favor of NantKwest, as we must, this evidence creates a genuine dispute of material fact that bars the grant of summary-judgment. Because the majority explains away NantKwest’s evidence instead of giving it the weight required by law, I respectfully dissent.
I.
As an initial matter, the difference between in vitro and in vivo testing is critical to understanding the difficulty in using results from the former to predict efficacy in the latter. In vitro experiments typically occur in the controlled environment of a petri dish or test tube; in vivo experiments are performed in a living organism. J.A. 886. Experiments in vitro cannot account for the variable environment of a living organism and cannot replicate a cell line’s interaction with the host’s immune system, among other things. J.A. 384, ¶ 24; J.A. 886. This disparity in testing environments can lead to unpredicted in vivo results. For example, cell lines with encouraging cytotoxic activity in vitro can unexpectedly lose all activity in vivo. J.A. 414, ¶ 93. The host’s immune system can even destroy the cell line, rendering it ineffective in vivo. Id. It is also possible for the cell line to trigger severe immune reactions in the host that produce serious complications. J.A. 393, ¶ 47. Therefore, demonstrated cytotoxic activity in vitro does not always translate to success in vivo.
II.
The presence of each claimed element in the prior art is insufficient to render a claim obvious. Rather, there also must be a motivation to combine the prior art and an ordinarily skilled artisan must have had a reasonable expectation of success in doing so. Kinetic Concepts, Inc. v. Smith & Nephew, Inc., 688 F.3d 1342, 1360 (Fed. Cir. 2012) (internal quotation marks and citations omitted). Whether a skilled artisan would have had a reasonable expectation of success is a question of fact, Cumberland Pharm. Inc. v. Mylan Institutional LLC, 846 F.3d 1213, 1222 (Fed. Cir. 2017), and contemporaneous evidence of what skilled artisans thought at the time of the invention can help inform our inquiry into whether the expectation of success was reasonable. In re Carroll, 601 F.2d 1184, 1186-87 (C.C.P.A. 1979). When determining whether a person of ordinary skill possessed a reasonable expectation of success in predicting in vivo efficacy based on in vitro results, we have recognized that “simply because a drug gives positive results in vitro, it does not necessarily follow that there is a reasonable probability of success for therapeutic use of that drug in vivo.” In re Gangadharam, 889 F.2d 1101, 1989 WL 127023, at *3 (Fed. Cir. 1989) (non-precedential) (citing Carroll, 601 F.2d at 1186).
The majority concludes that a person of ordinary skill in the art would have been motivated to combine Santoli with Gong because Santoli demonstrated in vivo efficacy for TALL-104 cells, Gong detailed NK-92’s in vitro ability to lyse cancer cells, and the prior art indicated a desire to seek clinical applications for the NK-92 cell line. Maj. Op. 867-69. NantKwest’s arguments that a person of ordinary skill in the art would not have had a reasonable expectation of success in combining Santoli with Gong were not persuasive to the majority in light of the “over-whelming specific evidence that a skilled artisan could reasonably extrapolate from the in vitro data with respect to TALL-104 and NK-92 that would reasonably teach their successful substitution in vivo.” Maj. Op. 870. In reaching its conclusion, the majority failed to give NantKwest’s evidence the weight it *876deserved and declined to draw all reasonable inferences in NantKwest’s favor.
Numerous contemporaneous references identified by NantKwest warned that in vitro results were not predictive of in vivo efficacy in this field. For example, the Vujanovic reference questioned the correlation between in vitro studies and in vivo behavior for the NK-92 cell line: “We suggest that standard in vitro cytotoxicity assays with target cells in suspensions have little relevance in predicting the in vivo antitumor activity of effector cells.” J.A. 962 (emphasis added). As confirmed by Dr. Miller, this passage from Vujanovic “caution[s] against relying on in vitro cyto-toxicity results, such as those in Gong, to predict in vivo behavior.” J.A. 468, ¶ 94. Yan contains similar warnings. Yan concluded his comparison of NK-92 and TALL-104 cells by noting: “[0]ur studies suggest that cytotoxic activity towards leukemic cell lines cannot be extrapolated to cells derived directly from patients. The use of such biologic reagents in vitro or in vivo for immunotherapy or purging must be accompanied by careful study of the unique patterns of activity....” J.A. 226 (emphasis added). Dr. Miller reiterated that “Yan admonishes against relying on in vitro tests to predict in vivo activity,” which was consistent with “the common understanding in the art.” J.A. 439-40, ¶ 48. Finally, in Cesano’s article describing a Phase I Clinical Trial for TALL-104 in dogs, the authors acknowledged that, “[sjurprisingly, the sensitivity of the dogs’ tumors to TALL-104 cell lysis in vitro did not appear to be a good indicator of clinical responses.” J.A. 963 (emphasis added).
NantKwest’s evidence and supporting expert testimony laid bare the uncertainty in this complex field. As Dr. Miller explained, the authors in the above references did not feel confident in predicting in vivo activity based on in vitro experiments at the time of the invention. J.A. 439-40, ¶ 48; J.A. 458, ¶ 94. This creates a dispute of material fact regarding the reasonable expectation of success. In my view, the majority’s willingness to discredit Dr. Miller’s understanding of the disclosures in Vujanovic and Yan does not dispose of the genuine dispute of material fact. Although the majority believes its view of the prior art references is superior to Dr. Miller’s, its analysis is not supported by citations to the USPTO’s expert report or the district court opinion. See Maj. Op. 873-74. Dr. Miller’s opinion, on the other hand, is illuminated by the background knowledge of a skilled artisan in this field, and I am not convinced that his opinion lacks support in the record. The result is a genuine dispute of material fact that I believe is not suited for resolution at the summary judgment stage.
Indeed, we have found lesser evidence sufficient to reverse rejections by the PTO in the past. In Carroll, for example, the PTO’s Board of Appeals rejected as obvious claims of a patent for treating M. paratuberculosis with lauric acid based on the patentee’s master’s thesis. The thesis disclosed two types of studies: 1) in vitro studies, from which the patentee reported that lauric acid completely inhibited the growth of three strains of M. paratubercu-losis, and 2) in vivo studies, from which the patentee reported the suitability of a certain strain of C57 black mice as laboratory animals for studying the diseases caused by M. paratuberculosis. Carroll, 601 F.2d at 1185. An expert in the field, Dr. Merkal, discounted the patentee’s thesis at the time of its publication because, among other reasons, “in vitro testing was an unreliable indicator for the in vivo effectiveness.” Id. at 1186. When the paten-tee sought a patent for his later discovery that mammals can be treated with lauric acid orally to treat M. paratuberculosis, the PTO argued that the claims were obvi-' *877ous because his earlier thesis disclosed lauric acid’s activity in vitro and that mice were suitable animals for studying the disease. Id. at 1185-86. Our predecessor court disagreed with the PTO, relying principally on Dr. Merkal’s “contemporaneous evaluation of appellant’s thesis[] that one skilled in this art would have given no weight to the findings reported therein.” Id. at 1186-87 (emphasis added).
Similarly, in Gangadharam, we held that the Board of Patent Appeals and Interferences erred in finding a reasonable expectation of success for using CQQ to treat tuberculosis in mammals (in vivo) based on in vitro results. Gangadharam, 1989 WL 127023, at *3 (nonprecedentia]). The Board relied on a single reference authored in part by the applicant to reject the claims. It found that the Gangadharam reference’s disclosure of “very positive in vitro bactericidal activity of CQQ against the [pertinent] bacteria reported by Gan-gadharam certainly favors the in vivo use of said compound in the treatment of tuberculosis in mammals.” Id. at *1. But simply “[rjemarking that the positive in vitro results ‘favored’ use in vivo does not meet the statutory standard,” we explained, and therefore the PTO “fell woefully short of its burden” to establish a reasonable expectation of success. Id. at *2. Importantly, the Gangadharam reference contained a proviso that further studies were needed “before CQQ can be suggested as a possible antimycobacterial drug for treating humans with [tuberculosis],” and a contemporaneous article warned that in vitro tests were neither equivalent to, nor a substitute for, in vivo experiments. Id. at *2-3. Because “there [wa]s evidence in this record ... regarding the noncorrelation of in vivo from in vitro efficacy generally and with respect to tuberculosis,” we found that the PTO failed to demonstrate a reasonable expectation of success. Id. at *3; see also Abbott Labs. v. Sandoz, Inc., 544 F.3d 1341, 1350-52 (Fed. Cir. 2008) (affirming non-obviousness in context of district court’s grant of preliminary injunction because, inter alia, the patentee’s evidence demonstrated that “it was not predictable” how an antibiotic would perform in vivo based on in vitro experiments using a closely related antibiotic).
As was the case with both Carroll and Gangadharam, the record here contains ample evidence that persons of ordinary skill in the art were skeptical about the ability of in vitro tests to predict in vivo efficacy for the two relevant cell lines. Vujanovic, Yan, and Cesano—in the very same articles where they discussed the promising in vitro results for the NK-92 and TALL-104 cell lines—proceeded to expressly caution against inferring efficacy in vivo based on these outcomes. A reasonable reading of these references supports the conclusion that one skilled in the art would not have had a reasonable expectation of success in combining Santoli’s in vivo testing for TALL-104 with Gong’s NK-92 cell line. Because NantKwest is the non-movant at the summary judgment stage, it is a reasonable inference that we must draw in its favor. See Anderson, 477 U.S. at 255, 106 S.Ct. 2505.
III.
Our standard of review for the grant of summary judgment requires us to believe the evidence of the non-movant and to draw all reasonable inferences in its favor. The majority did neither in finding the claims in NantKwest’s patent application obvious. Accordingly, I respectfully dissent.