Full Opinion

IN THE COURT OF CHANCERY OF THE STATE OF DELAWARE MINNIE SARWAL, ) ) Plaintiff/Counterclaim ) Defendant, ) ) v. ) C.A. No. 2023-0222-BWD ) NEPHROSANT, INC., ) ) Defendant/Counterclaim ) Plaintiff. ) POST-TRIAL MEMORANDUM OPINION Date Submitted: October 9, 2026 Date Decided: November 12, 2026 Michael A. Pittenger, Nicholas D. Mozal, Callan R. Jackson, Charles P. Wood, Ryan M. Ellingson, Lauren Angelina, POTTER ANDERSON & CORROON LLP, Wilmington, DE; Attorneys for Plaintiff/Counterclaim Defendant Dr. Minnie Sarwal. Thomas A. Uebler, Kathleen A. Murphy, Terisa A. Shoremount, MCCOLLOM D’EMILIO SMITH UEBLER LLC, Wilmington, DE; OF COUNSEL: Jeffrey D. Horst, Kana A. Caplan, Jessica G. Cino, KREVOLIN & HORST, LLC, Atlanta, GA; Attorneys for Defendant/Counterclaim Plaintiff NephroSant, Inc. DAVID, V.C. NephroSant, Inc. was founded to develop and monetize diagnostic tests derived from intellectual property developed in plaintiff Dr. Minnie Sarwal’s laboratory at the University of California, San Francisco School of Medicine. Those diagnostic tests included a non-invasive urine test known as “QSant,” used to identify the risk of kidney transplant rejection. A patient sends a urine sample to a lab that measures six “biomarkers” (cfDNA, m-cfDNA, clusterin, CXCL10, creatinine, and total protein) present in the sample, then the biomarker measurements are run through a machine learning algorithm to generate a numerical “score” between one and 100 correlating to the risk of kidney transplant rejection. The QSant test is a laboratory-developed test regulated by the Centers for Medicare & Medicaid Services through its Clinical Laboratory Improvement Amendments program. MolDX establishes coverage and reimbursement rates for diagnostic tests. To obtain coverage for QSant, NephroSant submitted a clinical dossier to MolDX that repeatedly described QSant as a six-biomarker test. NephroSant’s initial tests for QSant used frozen samples. As NephroSant began to test fresh samples, it determined that exposure to heat during shipping presented a risk that one of the biomarkers, clusterin, would degrade. Sarwal and NephroSant’s lab directors decided to exclude clusterin from the QSant algorithm, despite the company’s prior representations in MolDX submissions that QSant was 1 a six-biomarker test. Sarwal, who served as CEO and a board member, did not inform her fellow directors of this material change to the company’s lead product. A year later, after Sarwal stepped down as CEO, NephroSant employees raised concerns that the company’s lab had been running a five- instead of a six- biomarker QSant test. The board formed a special committee to investigate. After the special committee’s counsel directed Sarwal to refrain from deleting or altering company documents, Sarwal downloaded then deleted tens of thousands of documents from a company Box account. She also urged NephroSant investors to back her in a new competing venture, though those efforts did not progress. The board halted commercialization efforts, refunded Medicare reimbursements, and attempted to conduct a bridging study to validate the five- biomarker test. An initial validation analysis concluded that the five- and six- biomarker tests were not equivalent, and the company raised funds to pursue a new validation study, which it never completed. Instead, in May 2024, the company contributed substantially all of its assets to a new entity. This post-trial decision imposes sanctions for Sarwal’s document destruction. It concludes that Sarwal breached her fiduciary duty of loyalty by concealing material information about changes to the QSant test from the board, destroying company documents, and soliciting investors to form a competing business. It awards damages for those breaches, but nothing close to the $100 million 2 NephroSant’s damages expert sought to establish at trial. Finally, the decision finds that Sarwal is not entitled to indemnification in connection with the special committee’s investigation. I. BACKGROUND The following facts are as the Court finds them following a two-day trial held on May 26 and 27, 2026.1 A. Plaintiff Founds NephroSant Using Intellectual Property Developed At UCSF. Plaintiff Dr. Minnie Sarwal (“Plaintiff”) is a Professor in Residence of Transplant Surgery, is a Medical Director, and co-directs a Pancreas Transplant Program at the University of California, San Francisco School of Medicine (“UCSF”). 2 She holds a Ph.D. in molecular genetics from Cambridge University, a diploma in Child Health from University College London, and an M.D. from Calcutta Medical College. 3 In 2017, Plaintiff, with nonparties Josh Yang and Drew Watson, founded NephroSant, Inc. (“NephroSant” or the “Company”), formerly KIT.bio, Inc., with 1 The Joint Pre-Trial Stipulation and Order is cited as “PTO ¶ __”. Dkt. 299. Trial testimony is cited as “Tr. (Witness) at __”. Dkts. 303–04. Joint exhibits are cited as “JX __” unless otherwise defined. Dkt. 270. 2 PTO ¶ 42; Tr. (Sarwal) at 5:7–12. 3 Tr. (Sarwal) at 5:15–23; Pl.’s Opening Post-Trial Br. [hereinafter POB] at 2, Dkt. 311. 3 Plaintiff serving as Chief Executive Officer (“CEO”). 4 NephroSant was founded to develop and monetize diagnostic tests derived from intellectual property developed in Plaintiff’s “Sarwal Lab” at UCSF.5 Those diagnostic tests included a non- invasive urine test known as “QSant,” used to identify the risk of transplant rejection. 6 Because UCSF owns the intellectual property rights for innovations developed in the Sarwal Lab, NephroSant and UCSF entered into two agreements to facilitate NephroSant’s development of the QSant test. First, in February 2018, NephroSant and UCSF entered into an Exclusive License Agreement under which UCSF granted NephroSant an exclusive license to develop and market the QSant test.7 Second, in September 2019, NephroSant and UCSF entered into an Outgoing Data and Material Transfer Agreement (the “MTA”), which granted NephroSant access to frozen urine samples that Plaintiff had collected in the Sarwal Lab over 4 PTO ¶¶ 42–43, 67. 5 Id. ¶ 42; Tr. (Sarwal) at 8:22–9:7. 6 PTO ¶ 42. 7 Id. ¶ 70; JX 1. 4 twenty years. 8 NephroSant was to use the samples in its research, conduct additional studies to progress the research beyond reliance on the samples, and then return the samples to UCSF. The QSant test is intended to function as an at-home early-detection test that measures the risk of kidney rejection to enable early intervention.9 Broadly speaking, the QSant test is conducted in two stages—a “wet” side and an “algorithm” side.10 In the wet portion of the test, a patient sends a urine sample to the lab, where a test is conducted to measure six “biomarkers” present in the collected urine: cell- free DNA (“cfDNA”), methylated cell-free DNA (“m-cfDNA”), CXCL10, clusterin, total protein, and creatinine.11 In the algorithm portion, the biomarker measurements are run through a machine learning algorithm that employs a “random forest” model.12 The algorithm generates a numerical “score” between one and 100 (the 8 PTO ¶ 71; JX 2; Tr. (Sarwal) at 20:24–22:9. The MTA noted that the samples were covered by the Health Insurance Portability and Accountability Act of 1996 (“HIPAA”). JX 2 § 1. Separately, in August 2020, NephroSant and Plaintiff entered into a Confidential Information Agreement which stated that the data and intellectual property developed in connection with Plaintiff’s role at UCSF or developed with UCSF’s trade secret information during her professorship would remain UCSF’s property. JX 17 § 3. 9 PTO ¶ 43; Tr. (Sarwal) at 15:7–20. 10 Tr. (Ghosh) at 409:17–410:3, 410:16–411:3. 11 Id. at 409:17–410:3; JX 16 at 5–6. 12 Tr. (Ghosh) at 410:4–7. A random forest model is a machine learning tool that combines the output of multiple decision trees to reach a single result. See JX 439 at 21. 5 “Q-Score”) which correlates to the risk of kidney transplant rejection. 13 A higher Q-Score reflects higher risk; a Q-Score greater than thirty-two indicates kidney rejection. 14 To train the model, data scientists begin with the “ground truth”—samples from patients for whom kidney rejection or nonrejection outcomes are known.15 The machine learning model uses the “ground truth” to detect patterns in the biomarker measurements until it can accurately predict the risk of kidney rejection. 16 Once the algorithm is developed, it is “locked down” (so that it stops adapting to new data) before it can be applied to new samples in the next study. 17 Although Plaintiff contends that a laboratory-developed test remains subject to “constant improvements and amendments,” if a material change is made to a test, a lab must conduct a “bridging study”—a validation process used to prove that the updated test version will produce results equivalent to the prior test.18 13 Tr. (Sarwal) at 14:13–15:2. 14 Id. 15 Id. (Ghosh) at 410:16–411:19. 16 Id. 17 Id. at 411:20–412:6. 18 Id. (Sarwal) at 37:9–13; id. (Ghosh) at 439:20–23; id. (Stoll) at 466:19–467:3. 6 B. NephroSant Submits A Dossier To MolDX For CMS Coverage Describing A Six-Biomarker Test. The QSant test is a laboratory-developed test regulated by the Centers for Medicare & Medicaid Services (“CMS”) through its Clinical Laboratory Improvement Amendments (“CLIA”) program. 19 MolDX is a program administered by CMS that establishes coverage and reimbursement rates for diagnostic tests.20 Each submission is assigned a “Z-Code” to identify the test submitted to MolDX for reimbursement.21 By 2020, NephroSant had established a CLIA laboratory and hired three CLIA lab directors.22 NephroSant and UCSF also published a series of papers that provided clinical validation for the QSant test in NephroSant’s CLIA lab.23 As NephroSant prepared to make an application to MolDX, Plaintiff asked one of NephroSant’s regulatory consultants, Bruce Quinn, about “using 5 markers in the final CLIA Q[]Sant assay” to eliminate m-cfDNA after finding this biomarker 19 Id. (Sarwal) at 17:14–22. 20 PTO ¶ 84. 21 Tr. (Sarwal) at 34:9–12. 22 Id. at 32:4–7; Id. (Lai) at 256:7–17. 23 Id. (Sarwal) at 22:20–24. 7 “ha[d] little to no impact on performance of the assay.” 24 Quinn advised Plaintiff that “having a major 6 marker publication and a planned 5 marker CLIA version [wa]s a potential difficulty” and CMS would need to “give a different Z code for a 5 marker and 6 marker version.”25 After this exchange, NephroSant kept m-cfDNA as a biomarker in the QSant test.26 In June 2020, NephroSant submitted a clinical dossier to MolDX to obtain CMS coverage for QSant. 27 The dossier described NephroSant and UCSF’s clinical studies to support the clinical validity of QSant. First, it summarized a 2020 clinical study, the “Yang study,” that “developed and validated a score in urine for accurate detection of biopsy confirmed acute renal transplant rejection”: In the initial Q[]Sant foundational clinical study, we developed an algorithmic model using cfDNA and five additional biologically relevant biomarkers (m-cfDNA, Clusterin, CXCL10, Creatinine, and Total protein) to create a composite Q-Score to differentiate urine samples from patients with renal transplant quiescence (no rejection on a renal transplant biopsy) from transplant rejection (acute rejection on a renal transplant biopsy). We then validated the performance of the locked Q-Score for diagnosis of acute rejection in two independent 24 JX 3 at 2; see JX 421 at 1 (identifying Bruce Quinn as a “MolDx consultant who played a central role in the submission(s) the Company previously made to Medicare”). Plaintiff objects to JX 3 as hearsay, but this document shows Plaintiff was aware of a possible regulatory risk, not that m-cfDNA did not, in fact, impact the test. Joint Schedule of Evidence, Ex. A at 1, Dkt. 325; D.R.E. 801(d)(2). 25 JX 3 at 2. 26 JX 16 at 5. 27 PTO ¶ 85; JX 16. 8 datasets of biopsy matched urine samples from adult and pediatric renal transplant recipients, with and without acute rejection. . . . These samples were used to train, test and validate the performance of the Q-Score, a composite scaled algorithm (from 0-100) developed across all six biomarkers, to detect biopsy confirmed acute rejection with high sensitivity, specificity and accuracy. 28 Second, the dossier summarized another 2020 clinical study, the “Nolan study,” that “re-validated the assay for accurate detection of acute renal transplant rejection using the previously established threshold”: To improve the analytical assay rigor and quality, the research assays used in the Yang et al foundational study were transitioned to the NephroSant CLIA lab, validated as per CLIA guidelines . . . and the locked assays were utilized for a clinical validation study to re-validate the Q[]Sant assay and Q-Score . . . . The study included 223 independent biopsy matched urine samples obtained with informed consent from 215 pediatric and adult recipients of renal allografts, transplanted at one of three different transplant centers . . . . Of the 223 samples, acute kidney allograft rejection (AR) was confirmed by the paired biopsy read in 71 samples and a no-rejection or stable (STA) phenotype confirmed in 152 samples. Urine samples were collected from these patients from 1 to 1539 days post-transplant. Statistical methods and machine learning models were applied to generate a Q-Score, similar to the foundational clinical paper to evaluate the correlation between a phenotype of AR and the Q-Score threshold. . . . Random Forest modeling was used to identify the relationship between the different biomarkers, along with time post- transplant, for defining kidney transplant rejection or no-rejection 28 JX 16 at 14 (emphasis added); see Joshua Y. C. Yang et al., A Urine Score for Noninvasive Accurate Diagnosis and Prediction of Kidney Transplant Rejection, Science Translational Medicine, Mar. 2020. 9 status. The re-developed Q-Score from the training set, obtained using the Random forest model, was used to confirm if the threshold score of 32 still identified kidney allograft rejection. 29 Further describing NephroSant’s clinical tests, the dossier reiterated that: In the clinical validation study by Nolan et al . . . , we employed measurements of the six biomarkers (cfDNA, m-cfDNA, Clusterin, CXCL10, Creatinine, and Total protein) in a kidney transplant cohort which comprised of 71 AR and 152 of STA to re-train and re-validate the Q[]Sant test, now performed in the NephroSant CLIA Lab. This CLIA validation study was designed to conduct a comprehensive assessment of the Q[]Sant assay and the Q-Score on urine samples using the same biomarkers, same assays and same Q-Score analytics, as previously described in our clinical foundational paper could provide similar performance specifications, at a similar Q-Score threshold of 32, to differentiate AR from STA patients. . . . The utility of the individual biomarkers with the locked Q-Score and the locked threshold of 32 was applied to discriminate STA and AR outcomes by validation on 66 samples in the independent validation set. 30 On July 17, 2020, MolDX assigned a Z-Code for the QSant test. 31 C. The QSant Test Changes From A Six- To A Five-Biomarker Test. NephroSant’s initial tests for QSant used frozen urine samples collected from patients in the Sarwal Lab over a twenty-year period.32 The hope, however, was that the test could be commercialized as a diagnostic test on fresh samples taken from 29 JX 16 at 15 (footnote omitted) (emphasis added). 30 Id. at 15–16 (footnotes omitted) (emphasis added). 31 PTO ¶ 86. 32 Tr. (Sarwal) at 22:1–16. 10 patients at home.33 To test QSant on fresh samples, NephroSant enrolled patients in an early access program (“EAP”) from April through October 2021.34 One challenge NephroSant encountered in that process was that the chemical composition of urine rapidly degrades at room temperature, which can cause the biomarkers that QSant evaluates to lose their informative value. 35 NephroSant developed a preservative intended to slow degradation for three days so that the test could be taken at home and shipped to NephroSant’s CLIA lab via FedEx.36 But when NephroSant simulated extreme shipping conditions by exposing samples to high heat, the biomarker clusterin still presented a risk of degradation. 37 At some point after the MolDX application was submitted, during “weekly R/D meetings where the biomarker algorithm and biomarker stability studies were regularly examined” by Plaintiff and NephroSant’s CLIA lab directors, “a decision was made 33 Id. at 15:7–20. 34 Id. at 49:11–21. 35 Id. at 20:10–23. 36 Id.; JX 16 at 6 (“NephroSant has a urine preservative for stabilizing cfDNA and the other biomarkers that comprise the Q[]Sant test in urine. . . . This is essential to ensure stability of DNA during shipment of clinical specimens to NephroSant CLIA lab via FedEx overnight priority shipment.”); id. at 8 (“Addition of a urine preservative stabilized high, medium, and low patient samples concentrations for both Clusterin (n = 6) and CXCL10 (n =8) for at least three days with the CXCL10 samples stable even after six days at room temperature.”); id. at 9 (“Our results show that addition of a urine preservative stabilizes cfDNA and m-cfDNA, Clusterin, CXCL10, Total Protein and Creatinine in urine for at least 3 days at room temperature.”). 37 Tr. (Sarwal) at 45:15–46:6; JX 291 Interrog. 36. 11 to exclude a marker, Clusterin, to ensure assay robustness and reproducibility.”38 According to Plaintiff’s own interrogatory responses, “[t]his decision was based on the fact that this marker did not impact the Q[S]core, but exposed the QScore potentially to minor variations as it had a risk of marker degradation over very high heat . . . .” 39 Plaintiff, who served as a member of NephroSant’s board of directors (the “Board”), did not inform her fellow directors of this change to the QSant test. 40 Despite conceding in discovery responses that “a decision was made to exclude” clusterin from the test, Plaintiff claimed at trial that the test changed itself. According to Plaintiff, as samples were processed, the algorithm began to “evolve.”41 The wet portion of the test continued to collect data from six biomarkers, including clusterin. 42 But the machine learning algorithm evolved with the addition 38 JX 291 Interrog. 36. 39 Id. 40 PTO ¶ 93; JX 291 Interrog. 32. 41 Tr. (Sarwal) at 41:3–5; id. (Lai) at 262:16–263:2. See also JX 291 Interrog. 22: QSant has 6 biomarkers that are collected and analyzed. The algorithm is roughly a weighted average across the markers to generate the QScore. The QScore leveraged an AI based algorithm to generate the biomarker weights or coefficients. As more data was being collected, the AI-based algorithm evolved. This evolution was expected as presented in the MolDx submission and the CLIA lab certification. This evolution is expected to be further enhanced through the AQUA Registry as the AI-based algorithm is applied to additional patient data. 42 Tr. (Sarwal) at 42:10–16; id. (Ghosh) at 410:8–15. 12 of new samples until it “reached its optimal performance on five [biomarkers] and de-emphasized clusterin.”43 Contrary to Plaintiff’s “constant evolution” theory, NephroSant’s former Vice President of Data Science and Software Engineering, Dr. Srinka Ghosh, testified at trial that the algorithm was “locked” at five biomarkers by April 2021, despite NephroSant’s earlier representations to MolDX that it was employing a six- biomarker algorithm. 44 D. MolDX Approves A Six-Biomarker QSant Test For CMS Coverage. Throughout 2021, NephroSant made three updated submissions to MolDX, which continued to describe QSant as a six-biomarker test.45 MolDX approved the QSant test for Medicare coverage on August 20, 2021.46 After receiving MolDX approval, in September 2021, NephroSant launched a Series B financing round to raise $40 million.47 43 Id. (Sarwal) at 38:16–39:4, 43:9–11. 44 PTO ¶ 60; Tr. (Ghosh) at 435:20–23, 437:19. 45 PTO ¶¶ 87–89; JX 39 at 76; JX 42 at 7; JX 44 at 13. 46 PTO ¶ 90. 47 Id. ¶ 80. Plaintiff claims that at this time, the Company began storing documents in Box, a cloud-based storage service. Plaintiff created a folder in her NephroSant Box account “where [she] could keep some of [her] UCSF working files.” Tr. (Sarwal) at 90:5–92:3. 13 In February 2022, NephroSant published a new study based on the EAP samples entitled Through the Looking Glass: Unraveling the Stage-shift of Acute Rejection in Renal Allografts. 48 Plaintiff claims this paper “validated” the five- biomarker test, but it did not.49 Its methodology included collecting data from six biomarkers, and the paper did not address the effect (or non-effect) of clusterin in the algorithm portion of the QSant test.50 E. The Board Learns That The QSant Test Was Changed From A Six- To A Five-Biomarker Test And Forms A Special Committee To Investigate. On March 11, Plaintiff voluntarily stepped down as NephroSant’s CEO and Mark Capone replaced her as interim CEO.51 Under the terms of a Transition Agreement, Plaintiff assumed the role of Chief Scientific Advisor. 52 The following month, NephroSant employees raised concerns that the CLIA lab had been running a five-biomarker instead of a six-biomarker QSant test.53 Capone called a special meeting of the Board—comprising Plaintiff, Capone, Jeffrey 48 JX 64 at 1; id. at 2–3 (providing “a new analysis of a real-world data (RWD) cohort, comprising of prospectively collected urine samples (n = 235) from 11 different US transplant centers, collected between 1 April and 1 October 2021”). 49 Tr. (Sarwal) at 49:15–50:1. 50 JX 64 at 3. 51 JX 68 at 1. 52 PTO ¶ 42; JX 67 § 1(a)(i). 53 PTO ¶ 94. 14 Giullian, Michele Colucci, and Juan Loaiza—to discuss concerns about “QSant tests in the laboratory and the accuracy of the results reported by the Company.”54 Plaintiff emailed Board members, explaining she had heard that “some information was communicated by someone to somebody and perhaps to a board member that there was some information related with our Mol[DX] submission.”55 She warned that NephroSant was “at a very positive and critical inflection, with 6 weeks of cash, and [was] about to close a financing round,” and “[d]ecisions we make as a board at this point have to be held to a high fiduciary standard at this critical time for the company.”56 Separately, Plaintiff sent the Board a document summarizing the history of the QSant test, which included a chart identifying five versions of the test, including two (versions four and five) that showed clusterin was given a weight of zero. 57 At an April 16 special meeting, the Board established a special committee of directors comprising Giullian and Colucci (the “Special Committee”) “to investigate allegations related to the process for running QSant tests in the laboratory and 54 JX 82 at 1; PTO ¶ 94. 55 JX 79 at 1. 56 Id. 57 JX 81 at 4. 15 reporting the results” (the “Investigation”). 58 Plaintiff did not support formation of the Special Committee, expressing “concerns regarding the impact of a potential investigation on the Company’s cash runway and the ongoing Series B financing.”59 The Board nevertheless determined to “pause all QSant test reports until the Investigation reached a conclusion.”60 At the same time Capone brought the issue to the Board, he also asked Ghosh and another NephroSant employee, Francisco Dias Lourenco, to review the QSant algorithm.61 On April 16, Capone asked Lourenco, “[i]f we were going to do a proper validation” of the five-biomarker test, “how would we do that and how long would that take?”62 Lourenco responded: With the current set of samples we have from EAP we would probably have enough data to re-validate—but we don’t have permission to use the clinical data (we didn’t set up EAP to be a trial). As I mentioned, the QSant results highly correlate with the clinical picture of the patients from the feedback we receive. If we had to start collecting samples from scratch, the fastest I’ve seen a study done is three months —but it was with a well-established clinical affairs team at cepheid.63 58 PTO ¶ 94; JX 82 at 2. 59 JX 82 at 1. 60 PTO ¶ 95. 61 Tr. (Ghosh) at 417:3–15; id. (Sarwal) at 32:15–19 (identifying Lourenco as senior vice president of operations, product, marketing, and sales). 62 JX 84 at 2. 63 Id. at 1. 16 F. Plaintiff Deletes Documents From Her NephroSant Box Account. The Special Committee engaged Fenwick & West LLP (“Fenwick”) as its legal counsel in connection with the Investigation.64 On April 21, Fenwick emailed Plaintiff, requesting that she “provide the login credentials for [her NephroSant] Box account so administrative rights can be transferred over to the Company, and that [Plaintiff] refrain from deleting or altering in anyway the contents of the account.”65 Fenwick followed up with Plaintiff later the same day but did not receive a response. 66 On April 23, the Special Committee reported preliminary findings to the Board, including that “the QSant test remains effective, but the Company’s marketing and reporting materials need to be updated to reflect the composition and accuracy of the version of the QSant test the Company is currently deploying.”67 The Special Committee further reported “that the current version of the QSant test is not what is reflected in the Company’s prior submissions to MolDX, nor is it consistent with the descriptions on the reports.”68 The Special Committee recommended, based on its preliminary findings, that: 64 PTO ¶ 46. 65 PTO ¶ 97; JX 111 at 2. 66 JX 111 at 2. 67 JX 105 at 1. 68 Id. 17 (i) the Company’s Chief Executive Officer direct the efforts to review the algorithm utilized for the QSant test, and (ii) the Company continue to hold off on billing until further evaluation of the QSant test and, if appropriate, revision of the MolDX submission, reports, and marketing materials . . . . 69 After the Board meeting, Plaintiff asked a NephroSant employee to “write code” for a new six-biomarker algorithm.70 Plaintiff also worked with Lourenco on a presentation to the Special Committee proposing that NephroSant could either “keep [QSant] as 5-marker algorithm with internal documentation of product version evolution” or “[i]ncorporate clusterin to current algorithm to make it a 6-marker algorithm.”71 The Special Committee instructed Plaintiff to continue working from home and not to contact employees. 72 On April 25, Fenwick again requested that Plaintiff turn over her NephroSant Box account credentials.73 At trial, Plaintiff testified that she then spoke with Fenwick, who authorized her to delete HIPAA-protected patient information.74 She 69 Id. 70 JX 692 at 3 (Plaintiff instructing Yazar to create code on April 23, 2022, at 1:13p.m.). 71 JX 108 at 1, 3. 72 JX 107 at 1. 73 JX 111 at 1. 74 Tr. (Sarwal) at 98:11–99:10, 99:15–100:3. 18 further testified that Capone independently blessed her deletion of “UCSF proprietary” information. 75 Between April 21 and April 23—before Fenwick purportedly authorized the deletion of HIPAA-protected information—Plaintiff downloaded, deleted, and cleared from the “trash” folder tens of thousands of documents from her NephroSant Box account.76 Plaintiff also instructed her son, Reuben Sarwal (who also worked at NephroSant), to delete documents from her Box account, which he did.77 In total, Plaintiff “downloaded 33,000 documents and double deleted 15,000 of those.” 78 Plaintiff insists that she only deleted HIPAA-protected patient information and copies of documents that existed separately on the Company’s server.79 But—as discussed in further detail below—NephroSant claims that the deleted files contained data foundational to the Company’s research (the “Master Data File”).80 75 Id. at 97:21–98:8. 76 PTO ¶¶ 98–100; JX 690 at “BoxLogs” cells A32, A48395. 77 PTO ¶¶ 101–04. 78 Tr. (Colucci) at 196:4–6; see also JX 329 at 4 (expert report concluding that “15,304 items were either moved to the Trash folder, or were emptied from the Trash folder,” “33,008 files or folders were downloaded between April 19th, 2022, and May 16th, 2022,” and “11,640 files and folders were downloaded which no longer exist in Nephro[S]ant’s SharePoint environment”). 79 Tr. (Sarwal) at 103:17–24. 80 Id. (Stoll) at 473:22–474:4. 19 The Board placed Plaintiff on administrative leave and removed her access to the Company’s “digital and electronic systems” on April 26. 81 Plaintiff then engaged Orrick, Herrington & Sutcliffe (“Orrick”) to represent her in connection with the Investigation. 82 G. Plaintiff Meets With Investors About A Competing Business. In late May, Plaintiff met with NephroSant investors, including Haydar Alireza and William R. Magee, Jr., and suggested they back her in a “new venture” using NephroSant’s technology.83 Plaintiff told the investors that “if [NephroSant] goes into [bankruptcy], the IP will revert to UCSF, and then [Plaintiff] can start again.”84 Nothing came of these preliminary discussions. H. NephroSant Attempts To Validate The Five-Biomarker Test But Its Initial Validation Study Is Unsuccessful. On May 1, one of NephroSant’s CLIA lab directors, Dr. Donna Hongo, emailed NephroSant’s other lab directors and Lourenco, asking: “[f]or the current re-validation of the QSant algorithm, we are comparing to the original validation set 81 PTO ¶ 107; JX 110. 82 JX 666. 83 JX 299 at 4; Tr. (Magee) at 573:8–9. 84 Tr. (Magee) at 573:4–9. 20 that was use[d] in the Nolan et al paper and what was reported to CMS as our performance, correct?” 85 Copying Ghosh, Lourenco responded: The validation approach is to compare performance back to Nolan et al, and what was reported to CMS. For clarification, the current algorithm deployed to production (used in the lab), was also based on the same dataset as what we are doing with this new version. The issue with the current algorithm is that it does not have the same performance characteristics as the Nolan et al version, and uses 5 markers instead of 6. [Ghosh] is working with the team to correct both of those issues. And yes, we have access to the complete dataset. For clarification, the agreement with UCSF is with NephroSant, not [Plaintiff] specifically, so we still have access to data and the IP. The current plan is to complete validation of the algorithm this coming week, with a plan resumption of operations by Monday May 9th. 86 In this initial validation analysis (the “Initial Validation”), Ghosh looked at three sets of data—the “Nolan Data” collected from “Mexico, Stanford, [and] UCSF cohorts” “over the last 15 years”; the “Wanzin Data” collected from the “ARETAN” and “TITRATE” cohorts from 2018 through 2020; and the “EAP Data” collected in the EAP beginning April 2021.87 She compared the Nolan Data to the Wanzin Data and concluded that “26 samples [in the Nolan Data] which were originally diagnosed as nonrejection w[ere] changed to acute rejection in the second rendition of the data 85 JX 117 at 4; JX 633 at 2; Tr. (Sarwal) at 32:11–12. 86 Id. at 3 (emphasis in original). 87 JX 601 at 4. 21 cohort.”88 Ghosh believed someone manually altered the data to change some patients’ diagnoses. 89 Ghosh also sought to “[d]etermine the effect of rescoring of EAP samples” against a new six-biomarker test.90 Ghosh concluded that the five- and six-biomarker tests were not equivalent. 91 Ghosh presented her findings to Capone, and the Board ultimately decided to “cease [NephroSant’s] commercial activities” and “conduct a prospective validation of the QSant test.”92 On May 10, Capone resigned as a director and interim CEO of the Company, and Colucci was named CEO. 93 The following week, the Company sent a letter to Fresenius Medical Ventures GmbH—which had offered two weeks prior to invest $10 million in the Series B financing round—disclosing the Investigation. 94 The letter explained that “material changes in the specifics of the QSant test led . . . experts to conclude that reliance on a QSant value obtained during real world testing 88 Id. at 20; Tr. (Ghosh) at 418:20–23; see also JX 601 at 6 (“26 samples changed from NR in Nolan to AR in Wanzin”). 89 Tr. (Ghosh) at 419:10–12; id. at 419:5–8 (“Q. And were you able to determine from your data review who changed the phenotypes from nonrejecting to acute rejection? A. It was Dr. Sarwal.”). 90 JX 601 at 7. 91 Id. at 16; Tr. (Ghosh) at 422:6–423:4. 92 JX 125 at 1–2; JX 601; Tr. (Ghosh) at 417:10–15. 93 PTO ¶ 110. 94 PTO ¶ 53. 22 was potentially unwarranted and that there should be a resetting and a reframing of the algorithm on which QSant is based.”95 Fresenius subsequently withdrew its offer.96 Later that month, Plaintiff and NephroSant’s CLIA lab directors made a presentation to the Board to justify the five-biomarker algorithm and propose steps to return to commercialization.97 One slide conceded that there was “[n]o [c]lusterin” in the “[c]urrent CLIA [d]eployment.”98 I. The Special Committee Concludes Its Investigation And Plaintiff Resigns As Chief Scientific Advisor. From April to June, the Special Committee conducted approximately forty- three interviews, including three interviews with Plaintiff.99 On June 14, the Board held a meeting at which the Special Committee presented the following “summary recommendations” based on its Investigation: [T]he Company should (i) engage regulatory experts and counsel to advise on remediation and disclosure, if any, including to Centers for Medicare and Medicaid and MolDX, (ii) develop standardized processes and clear documentation consistent with scientific and clinical laboratory best practices; consider oversight by independent third party, (iii) clearly identify roles and responsibilities, (iv) create independent processes for future validation efforts, (v) enforce culture 95 JX 140 at 3. 96 PTO ¶ 116. 97 PTO ¶ 93. 98 JX 605 at 5. 99 PTO ¶¶ 94, 119. 23 of openness, not “siloed” information, (vi) conduct training and set clear expectations for management of issues that must be escalated to the Board and (vii) evaluate personnel decisions . . . . 100 In total, the Company incurred $827,682.84 in legal fees in connection with the Investigation.101 One month later, on July 13, Plaintiff resigned as Chief Scientific Advisor and made a demand for unpaid compensation and indemnification of expenses incurred in connection with the Investigation. 102 Plaintiff resigned from the Board on July 31.103 J. NephroSant Pursues The PARK Study, But Then Transfers Its Assets To KidneyMetrix. After the Investigation concluded, NephroSant continued to evaluate the QSant test. NephroSant’s Vice President of Product Management, Jeffery Stoll, was tasked with locating the data required to support a formal validation of the five- biomarker test.104 At trial, Stoll testified that the Company lacked sufficient data to 100 JX 178 at 1–2. 101 JX 604. At the pre-trial conference, the Court granted Plaintiff’s motion in limine to exclude belatedly produced evidence purporting to summarize the Special Committee’s findings. See JX 191; Tr. of Pretrial Conference [hereinafter PTC Tr.] at 29:19–24, Dkt. 326. 102 PTO ¶¶ 121–22; JX 246. 103 Tr. (Sarwal) at 87:19–24. 104 Id. (Stoll) at 464:16–23, 468:10–21. 24 validate the test results because it could not recover the “Master Data File” Plaintiff purportedly deleted in April 2022. 105 In August, the Company raised $9.27 million in bridge financing to pursue a new validation study (the “PARK Study”).106 The “[k]ey [o]bjective” of the PARK Study was to “[u]pdate QSant[’s] 6 biomarker algorithm and validate performance specs” using data from nine transplant centers from 500 samples collected over six to nine months.107 NephroSant never finished the PARK Study. 108 NephroSant informed MolDX that the PARK Study “[r]equire[d] minimal data collection” and that 105 Id. at 469:5–17, 472:17–473:3. 106 Id. (Colucci) at 203:20–23, 206:14–17. On July 9, 2022, NephroSant executed a term sheet with a group