Full opinion text
OPINION & ORDER KATHERINE B. FORREST, District Judge: On March 11, 2014, Regeneron filed twin patent infringement actions: one against Merus B.V. (“Merus”), a company based in the Netherlands, and another against Ablexis LLC (“Ablexis”). (See ECF No. 1.) In short complaints, each consisting of a few substantive paragraphs, Regeneron accused both companies of infringing U.S. Patent No. 8,502,018 (“'018 Patent”). Me-rus answered and counterclaimed, arguing that the '018 Patent was unenforceable due to Regeneron’s conduct during patent prosecution. (See ECF Nos. 72, 225.) This litigation ensued. Following issuance of this Court’s opinion on claim construction, (ECF No. 210) Regeneron stipulated that its infringement claim as to Merus must fail if the Court’s constructions withstand challenge on appeal. (ECF No. 271.) Thereafter, Ablexis settled with Re-generon prior to claim construction, all that remained was Merus’s counterclaim for inequitable conduct. On June 9-15, 2015 the Court held a bench trial on that claim. Set forth below are the Court’s findings of fact and conclusions of law. Based on substantial evidence adduced at trial — as well as certain instances of Regeneron’s litigation conduct — it is clear that this litigation should never have been commenced. It is not unusual for one litigant to argue as much at the outset of a case, but it is much rarer for the evidence to prove it to be true. It is true here. Throughout the history of this case Regen-eron has sought to discover how it needed to define its invention to have it fit a cognizable theory of infringement; it has had to contort science, the documentary record, and an alleged commercial embodiment to make them fit the framework of a specification that described a far broader, not as useful, and possibly altogether different invention; and it has demonstrated that the invention disclosed in the '018 Patent is not the same as that Regeneron described during prosecution to the U.S. Patent & Trademark Office (“PTO”). As it turns out, the invention that Regener-on’s technical expert, Marjorie A. Oettinger, Ph.D., described is interesting and might in fact lead to the discovery of therapeutically useful antibodies, but it is simply not the invention disclosed in the '018 Patent. It is unfortunate that this case has been marked by troubling litigation tactics, and doubly so as the purpose of this final proceeding was to determine whether Regen-eron had engaged in inequitable conduct or affirmative egregious misconduct during patent prosecution. Troubling litigation tactics were on display soon after this case was filed and continued into the trial. Based upon the Court’s assessment of the evidence, it appears that the very birth of this patent was beset by misconduct as well. And so it has come full circle. That which was obtained by misconduct ends as a result of misconduct. I. THE '018 PATENT Regeneron’s '018 Patent is the subject of this proceeding. Entitled “Methods of Modifying Eukaryotic Cells,” it is one of a number of patents .and/or related patent applications in the same family and sharing some or all of the same specification. (See '018 Patent, “Related U.S. Application Data.”) The original application to which the '018 Patent traces back was initially filed on February 16, 2001. (Id.) As discussed in several parts of this Opinion, this date- — February 16, 2001 — plays an important role in defining the invention; that is, in determining what it is and what it simply cannot be. - Regeneron describes the invention disclosed in the Patent as a mouse with normal immune response useful for discovering therapeutic antibodies. According to Regeneron, mice described by prior art had deficient immune response. The invention involves, in part, the targeted insertion of unrearranged human variable region DNA segments into an endogenous mouse (murine) immunoglobulin (“Ig”) locus. , According to Regeneron, this would result in a mouse with human variable regions and mouse constant regions, that is, a “chimeric” or “reverse chimeric” mouse. Notably, and as described further below, Regeneron’s view of the invention necessarily presumes a multi-step process. The process could unfold in two different ways. It could be achieved by making two targeted insertions into the same mouse Ig loci, one of human heavy chain variable" regions and a subsequent and further targeted insertion of human light chain variable regions. Or, alternatively, it could be achieved by initial insertion of heavy and light chain variable regions into two separate mice and the subsequent breeding of the two mice, resulting in a mouse with both human heavy and light chain variable regions. An aspect of this targeted insertion is, according to Regeneron, placement at a precise point: the human variable region gene segments must be adjacent to the mouse constant regions. Regeneron’s technical expert, Dr. Oettinger, refers to this as “functional” linkage. In addition, Regeneron asserts that a necessary part of this invention includes' retention of mouse regulatory regions, specifically the tran-smembrane and cytoplasmic tail. Regen-eron asserts that the commercial embodiment of the invention is its Veloclmmune mouse. These aspects of the invention are important to the issues here before the Court. A differently defined invention runs directly into the prior art Merus claims Regeneron failed to disclose during patent prosecution. According to Merus, the invention (and claim 1 in particular) does not contain all of the limitations Regeneron now asserts. As an initial matter, it is not limited to mice with entirely human variable regions and entirely mouse constant regions, but may include those with combined human and mouse variable regions (that is, there is an insertion of human variable, but no deletion of the mouse, or insertion of human heavy chain leaving the mouse light chain, or vice versa). In addition, according to Merus, although claim 1 specifies that the insertion must occur into the Ig locus, it does not require insertion at the particular point within the locus (adjacent to, but neither overlapping with nor short of, the mouse constant region) as Regener-on now asserts; and this lack of specificity could lead to a mouse with an impaired immune response. Finally, according to Merus, the Veloclmmune mouse, which Regeneron represented to the PTO was the commercial embodiment of the '018 Patent, did not exist in February 2001; Regeneron only succeeded, in making it several years later, after a number of failed attempts — and then by a process different from that disclosed in the '018 Patent. II. DESCRIPTION OF THE TECHNOLOGY IN THE PATENT According to the specification of the '018 Patent, the method used to engineer this chimeric mouse involves “utilizing large DNA vectors to target, via homologous recombination, and modify, in any desirable fashion, endogenous genes and chromosomal loci in eukaryotic cells.” (’018 Patent, “Abstract.”) The large DNÁ vectors used in this process are defined in the specification as “LTVECs” — that is, large targeting vectors for eukaryotic cells. LTVECs are “derived from fragments of cloned genomic DNA larger than those typically used by other approaches intended to perform homologous targeting in eukaryotic cells.” (Id., 9:39-42.) A “targeting vector” is “a DNA construct that contains sequences .‘homologous’ to endogenous chromosomal nucleic acid sequences flanking a desired genetic modifi-eation(s). The flanking homology sequences, referred to as the ‘homology arms’, direct the targeting vector to a specific chromosomal location within the genome ...” (Id., 8:66-9:4.) All of the figures in the specification show versions of homologous recombination with LTVECs of various types and sizes, none smaller than 20 kilobases (“kb”). For instance, Figures 1 and 2 in the specification show a DNA “modification cassette” (or insert) being transferred by homologous recombination into a large targeting vector in a mouse’s genome: (Id., Fig. 1.) (Id., Fig. 2.) Figures 4A-4D of the '018 Patent show a human insert in a LTVEC of 200-300 kbs: Vs Da Jk eonsiaat region .. —rt -•8mBSeasanH«®iawBswi8»ii»BBMiiiSBBaBVSRa-sg«iB»8Ba- — MiiiBa—mwmsí ¡roman insect in UVEO additional lmnvm ssaastines human insert it) WVItCJ (-200 -308 M>) (-200-300 M>) ligate 4B Mosses lO heavy drain locus (loial length »lMb, nor drawn 10 scale): Vs Ds Is constant region -..........¡«i» — mmw-ms-ma-mouse tomolpgy ante in LTVEC2 «roasts homology antra mI„TVECl Figure 4C LTVEC2: Figure 0 LtVliCt: The specification states that “use of LTVECs provides substantial advantages over current methods” of homologous recombination. (Id., 1:37-38.) “LTVECs can be more rapidly and' conveniently generated from available libraries of large genomic DNA fragments (such as BAC and PAC libraries) than targeting vectors made using current technologies.” (Id., 1:40-43.) “The present' invention” is described as providing for “a rapid, convenient, and streamlined method for systematically modifying virtually all the endogenous genes and chromosomal loci of a given organism.” (Id., 1:51 — 54.) The specification also states that “[i]n accordance with the present invention, Applicants provide novel methods that enable the use of targeting vectors containing large regions of homology so as to modify endogenous genes or chromosomal loci in eukaryotic cells via homologous recombination. Such methods overcome the [limitations in the prior art.]” (Id., 2:64-3:2.) In the “Summary of the Invention,” the specification states, “In accordance with the present invention, Applicants have developed a novel, rapid, streamlined, and efficient method for creating and screening eukaryotic cells which contain modified endogenous genes or chromosomal loci.” (Id., 3:11-14.) The method uses LTVECs, introduces them into eukaryotic cells to modify the endogenous chromosomal locus of interest, and analyzes the cell with an assay for modification of the allele (“MOA assay”). (Id., 3:15-25.) The '018 specification references thirty (30) preferred embodiments, including several which are far broader than the invention Regeneron now describes. For instance, • “A preferred embodiment of the invention is a method for genetically modifying an endogenous gene or chromosomal locus in eukaryotic cells [using LTVECs].” (Id., 3:27-30.) • ‘Yet another preferred embodiment is a genetically modified eukaryotic cell that is produced by the method of the invention.” (Id., 4:6-8.) Certain embodiments make clear that deletions and insertions of genetic sequences are separate and distinct concepts, and that one is not assumed within the other (that is, an insertion of a genetic .sequence does not necessarily imply a deletion of a sequence). For instance, • “Another embodiment of the invention is a method wherein the genetic modification to the endogenous gene or chromosomal locus comprises deletion of a coding sequence, gene segment, or regulatory element; alteration of a coding sequence, gene segment, or regulatory element; [or] insertion of a new coding sequence, gene segment, or regulatory element ...” (Id., 3:40-43 (emphasis added).) • “An alternative embodiment of the invention is a method wherein the alteration of a coding sequence, gene segment, or regulatory element comprises a substitution, addition, or fusion ...” (Id., 3:48-51 (emphasis added).) • “Also preferred is a transgenic mouse having a genome comprising entirely human heavy and light chain variable region loci operably linked to entirely endogenous mouse constant region loci such that the mouse produces a serum containing an antibody .. \; a transgenic mouse containing an endogenous variable region locus that has been replaced with an homologous or ortho-logous human variable locus, such mouse being produced by a method comprising [obtaining and using LTVECs].” (Id., 7:24-38 (emphasis added).) Other embodiments directly contradict Regeneron’s description in this proceeding of its invention as requiring a mouse with entirely human variable regions and entirely mouse constant regions. For instance, • “One embodiment of the invention is a method of replacing, in whole or in part, in a non-human eukaryotic cell, an endogenous immunoglobulin variable region gene locus with an homologous or orthologous human gene locus comprising [using the LTVEC process to] introdúcete] the LTVEC ... into the eukaryotic cells to replace, in whole or in part, the endogenous immunoglobulin variable gene locus ... ”. (Id., 5:44-60 (emphasis added).) • “Another embodiment of the above method is a method wherein [certain steps] are repeated until the endogenous immunoglobulin variable region gene locus is replaced in whole with an homologous or orthologous human gene locus.” (Id., 6:11-15 (emphasis added).) • “Another embodiment of the method is one in which the immunoglobulin variable gene locus is a locus selected from the group consisting of a) a variable gene locus of the kappa light chain; b) a variable gene locus of the lambda light chain; and c) a variable gene locus of the heavy chain.” (Id., 6:16-20 (emphasis added).) • “Also preferred is an embryonic stem cell wherein the mouse heavy chain variable region locus is replaced, in whole or in part, with a human heavy chain variable gene locus; an embryonic stem cell of claim wherein the mouse kappa light chain variable region locus is replaced, in whole or in part, with a human kappa light chain variable region locus; an embryonic stem cell wherein the mouse lambda light chain variable region locus is replaced, in whole or in part, with a human lambda light chain variable region locus; and an embryonic stem cell wherein the heavy and light chain variable region gene loci are replaced, in whole, with their human homologs or orthologs.” (Id., 7:6-18 (emphasis added).) • “Yet another preferred embodiment is an antibody comprising a human variable region encoded by the genetically modified variable gene locus of described above; an antibody further comprising a non-human constant region; and an antibody further comprising a human constant region.” (Id., 7:19-23 (emphasis added).) • “Also preferred is a transgenic mouse having a genome comprising entirely human heavy and light chain variable region loci operably linked to entirely endogenous mouse constant region loci such that the mouse produces a serum containing an antibody comprising a human variable region and a mouse constant region in response to antigenic stimulation; a transgenic mouse having a genome comprising human heavy and/or light chain variable region loci operably linked to endogenous mouse constant region loci such that the mouse produces a serum containing an antibody comprising a human variable region and a mouse constant region in response to antigenic stimulation ...” (Id., 7:24-35 (emphasis added; note inclusion of the word “entirely” in the first example and the absence of that word in the second).) Notably, the preferred embodiments are consistent with Merus’s description of the invention and, as described below, the broadest reasonable construction of the claims. A. Certain Technical Principals Relevant to this Opinion Technical experts retained by Regener-on and Merus testified at the trial: Marjorie A. Oettinger, Ph.D. for Regeneron and Geoff Davis, Ph.D., for Merus. Both have substantial expertise in their fields. The Court relies upon both in its findings with regard to various basic technical principles relevant to the issues before the Court. Where one differed from the other in a material way, several examples of which are described in this Opinion, the Court found Dr. Davis more persuasive and his opinions based on a more substantial foundation. All technical determinations in-eluded in this Opinion constitute findings of fact. The technology relating to the '018 Patent generally involves a method to genetically modify mice to contain large fragments of human genomic DNA by use of targeting vectors (described below) and assay. The goal is to produce antibodies useful in drug discovery and, eventually, production of potentially useful therapeutic antibodies. (Davis Tr. Decl. ¶ 17.) DNA is a molecule in a cell which carries the genetic material for living organisms and is capable of self-replication and synthesis. It consists of a double-stranded molecule that pairs in a double-helical structure: “One end of each strand is called the 5-prime (5’) end, and the other is called the 3 prime (3’) end.” The 5-prime and 3-prime ends define the boundaries of a strand of DNA. One of the issues before the Court— and which was also at issue during claim construction — is whether the metes and bounds of the 5' end of the immunoglobu-lin locus was sufficiently understood as of the relevant date, February 16, 2001. The concern is whether practicing the invention required defining the 5' end of the locus, which is the beginning of the variable region. Merus asserts that without such definition, targeted insertion of a human variable region gene segment could miss the locus altogether, or it could fall short of insertion at a point that would reliably produce a useful antibody. Regeneron, in contrast, asserts that while the precise metes and bounds of the 5' end were not known in 2001, enough was known as to the size of the loci to allow for practice of the invention. As discussed below, the Court is persuaded by Merus’s expert Dr. Davis on this point and disagrees with Regeneron. DNA molecules are made up of chemical building blocks called “nucleotides”. Nucleotides on the two strands of the double-helix pair with one another in complementary units called “base pairs.” The base pairings connect the individual DNA strands to one another to form the double-helix. The unique sequence of bases on a given strand represents a code; a gene is a unit of DNA that includes the sequence of bases representing the codes for the amino acids that comprise a particular protein. In this case, the concept of kilobase pairs has some relevance. The term “kilo-base pairs” refers to a DNA strand 1,000 base pairs long. In the '018 Patent, a core aspect of the invention is the utilization of a large fragment of DNA — measured in kilobase pairs — for targeted insertions. Genes are expressed by cells as proteins through processes commonly referred to as “transcription” and “translation”. Before transcription and translation, the two strands of DNA that constitute a gene unwind from their double-helix configuration. During transcription, machinery in the cells reads the DNA sequence of one of the DNA strands, nucleotide by nucleotide, and uses it as a template to produce an intermediate molecule called messenger RNA (abbreviated as mRNA). The structure of a protein gives rise to its biologic activity. A key benefit of the invention as described in the '018 Patent is successful B cell replication. “B cells” make antibodies. Antibodies are also known as “immunoglo-bulins.” They are a particular type of protein with the potential to bind specifically to foreign antigens. (Oettinger Trial Aff. ¶ 22.) All immunoglobulins have a similar structure. (Id. ¶ 23.) They are typically depicted as having a structure shaped like the letter “Y”. (See id. Fig. 1.) The Y structure consists of four chains of amino acids: two identical light chains and two identical heavy chains. Each light chain pairs with a partner heavy chain, and then each heavy-light chain pair associates with an identical heavy-light chain pair to form the “Y” structure. See Figure be-jow. Each Ig heavy or light chain is composed of several regions. (Oettinger Trial Aff. ¶ 24.) Within an immunoglobulin subunit, there are regions with extensive amino acid sequence variations between them, which are called “variable” regions. (Id.) Regions that show no sequence variation within a species are called “constant” regions. (Id.) Each heavy chain and light chain is comprised of a “constant” region and a “variable” region. In both heavy chains and light chains of an antibody, the region at the tip of the “Y” is the variable region. The other region on each heavy chain and light chain is the constant region. See Figure 2: The heavy chains are referred to by the letters of the Greek alphabet. The heavy chains of the different classes of immuno-globulins, IgM, IgD, IgG, IgA and IgE, are referred to as (mu), S (delta), á (gamma), á (alpha) and c (epsilon), respectively. (Id. ¶27.) Oettinger states that “a comparison of the constant regions between different species ... reveals important differences. For example, although the amino acid sequences of the constant regions of mouse and human IgGl have about 70% sequence identity, there are numerous amino acid substitutions that distinguish one from the other. These species — species features are important when considering the functional and antigenic properties of human immunoglobulins introduced into a different species ...” (Id.) Important to the issues before this Court is the fact that, as Dr. Oettinger testified and as further set forth in Dr. Davis’s Trial Declaration (See Davis Tr. Decl. ¶ 322), replacement of an entire variable region requires at least two steps: replacement of the heavy chain of the variable region, and replacement of the light chain. (ECF No. 398, 674:2-19.) Insertion of an entire exogenous variable region requires two similar insertion steps — one for the heavy chain and one for the light. It would not matter which order such replacement or insertion was accomplished— but it cannot be accomplished in a single step. (Id., 674:20-675:2.) Antibodies are proteins composed of amino acids, encoded by genes composed of DNA nucleotides. The DNA that encodes antibody variable regions is assembled from separate gene “segments.” A gene that encodes the heavy chain variable region of an antibody is assembled from three gene segments, named the variable (V or VH), diversity (D or DH) and joining (J or JH) segments (the subscript “H” indicates the gene segment that forms part of the antibody heavy chain). A gene that encodes the light chain variable region of an antibody is assembled from two gene segments, named the variable (V or VL) and joining (J or JL) segments (the subscript “L” similarly indicates the light chain). These gene segments are joined together to form contiguous variable region gene segments (V(D)J for heavy chains, and VJ for light chains) through DNA rearrangement mechanisms. (Oet-tinger Tr. Aff. ¶35.) The genetic structure of the immunoglobulin loci together with the capacity of immunoglobulin DNA to (1) rearrange, (2) switch, and. (3) further mutate, allows for the production and development of a diverse antibody repertoire; these activities may also be referred to generally as recombination, isotype switching, and hypermutation. (Davis Tr. Decl. ¶ 22.) In order to generate mice that produce humanized antibodies, the '018 Patent sets out a method of integrating human genomic immunoglobulin DNA into the mouse genome. (Id. ¶ 23.) In both humans and mice there is one gene locus containing the genetic material used for expressing heavy chains, and two gene loci containing genetic material used for expressing light chains. Through a process known as V(D)J recombination, the DNA sequence encoding a variable region of an antibody heavy or light chain is created at each Ig gene locus by selecting and joining together one each of the many V, D and J gene segments (for heavy chains) or V and J gene segments (for light chains) present at the locus. (Oettinger Tr. Aff. ¶ 41.) V(D)J recombination is referred to as “somatic recombination”. See Figure 3: V(D)J recombination (ie., somatic recombination) is part of the process of B cell development essential to encode a complete antibody. All antibodies made by one B cell are identical. (Id. ¶ 29.) Thus, in order to have a diversity of antibodies, a diversity of B cells is required. B cell rearrangement is essential to that process. Somatic mutations (ie., changes in DNA sequences in B cells as opposed to germline cells) then further act to increase the affinity of an antibody with a given specificity. (Id.) “B cells arise in the bone marrow, where lymphoid progenitor cells develop into ‘immature’ B cells.” (Id. ¶ 30.) During this developmental process, rearrangements take place in the immuno-globulin genes. This is the “V(D)J recombination” discussed above. (Id.) If a successful gene rearrangement takes place, the B cell eventually acquires the capacity to display a “B cell receptor” on its surface; this B cell receptor is a “membrane-bound version of an immunoglobulin.” (Id.) Successful rearrangement of the heavy chain locus allows the B cell to produce a membrane-bound (mu) chain. (Id. ¶ 43.) In early B cell development, the membrane-bound (mu) chain is the first functional Ig subunit that is expressed: at this time, the light chain genes have not yet rearranged and the B cell does not make the complete B cell receptor, so the B cell is not capable of specific antigen recognition. (Id.) At this stage, only a membrane-bound (mu) heavy chain is expressed and the cell that carries it is referred to as a “pre-B cell.” (Id.) The membrane-bound (mu) chain is anchored in the membrane of the pre-B cell with the bulk of its mass facing outward. (Id. ¶ 44.) The membrane-bound (mu) chain forms a complex with two so-called “surrogate” light chains and two “accessory proteins;” this complex is referred to as the pre-B cell receptor (“pre-BCR”). (Id. ¶ 45.) The |x (mu) region, which has both a transmembrane domain and a cytoplasmic tail, plays an important role in this proceeding. Regeneron asserts that a benefit of the invention in the '018 Patent is retention of a mouse |i, (mu) region when human (heavy and light chain) variable regions have been inserted. Merus argues— and Davis persuasively supports this argument — that nothing in claims 15 requires retention of the murine |x (mu) region. There are — and at the time of the invention in February 2001 there already were — numerous methods for incorporating exogenous DNA, such as human DNA, into mice. The first method was the insertion of a “transgene” by random integration. A “transgene” is a DNA sequence originating from outside the host organism. One may create a “transgenic mouse” by injecting an exogenous DNA fragment into a fertilized mouse egg. (See ECF No. 400, 850:10-20; Davis Tr. Decl. ¶¶ 19, 59.) The DNA fragment is then incorporated into a random chromosomal location in the genome of the embryo. The exact location where the transgene DNA ends up in the genome is random. This process is known as “random integration”. Random integration may result in the location of the added DNA in an area which is more or less transcriptionally active and can also disrupt or render nonfunctional DNA regions into which it integrates. In other words, the inserted DNA may or may not be where you want it to be in the mouse genome. Non-randomized methods of genetic modification also existed prior to February 2001. The methods are sometimes generally referred to as “gene targeting.” (Davis Tr. Decl. ¶ 19.) The key technique used for targeted gene modification that had been developed before 2001 is called “homologous recombination.” (See, e.g., id. ¶¶ 18-20.) Homologous recombination relies on a vector (one can think of this as a “chunk”) comprised of the foreign DNA that one seeks to insert, flanked by regions of DNA that are homologous to the desired integration site, known as the homology arms. (Id. ¶ 19.) To facilitate homologous recombination, the DNA sequence of interest is flanked by “homology arms;” these arms consist of DNA fragments that are substantially the same in sequence as the sequences that flank the target DNA sequence being replaced or augmented in the genome. These arms allow the targeting construct to alight with the host genome to ensure modification at the desired position. Targeted insertion directs the DNA from the vector to integrate at a particular , site (or location) without changing the nearby regions of the host genome (this is an “insertion”), or it can direct the foreign DNA to replace a portion of the host genome with the foreign DNA to be integrated (“replacement” or “substitution”); this may include a deletion step. (Id.) An example of targeted insertion — -without deletion — is shown below in Figure 4 as integrated DNA without removing the DNA of the targeted genome: Insertion without deletion differs from “replacement” or “substitution” of mouse DNA with homologous human DNA. Implicit in replacement or substitution is the concept of removal, or possibly other inactivation, of the original gene segment. In this way, the mouse DNA would not be present (or active) in the mouse cell and the human DNA would be. (Davis Tr. Decl. ¶ 19.) The '018 Specification teaches a method of homologous recombination between a mouse and human in which the specific target is the immunoglobulin locus. In the '018 Patent, the mouse is the host, and a portion of an homologous human gene segment (here, some or all of the variable region) is inserted into the mouse’s immu-noglobulin locus. To do this, a “targeting vector” must be created. As described above, a “vector” is a vehicle which holds the DNA sequence (or gene segment) that the scientist intends to be incorporated into the mouse genome. To facilitate homologous recombination, the DNA sequence of interest is flanked by “homology arms”; these, again, are DNA fragments that are substantially the same in sequence as the sequences that flank (or are at either end) of the target DNA sequence being replaced or augmented in the genome. These arms allow the targeting construct to align with the host genome to ensure modification at .the desired position. Put otherwise, to drop the gene sequence into a particular locus you need a beginning and end that matches the beginning and end of the same sequence in the host; and to review (because this is complicated stuff) the entire segment is called the vector and the beginning and ends are the “arms” or “homology arms” of that vector. When the amino acid sequence of a protein is represented in a linear fashion, it is represented by convention with the “amino-terminal” end on the left and the “car-boxyl-terminal” end on the right. (Oet-tinger Trial Aff. ¶28.) For nucleic acids such as DNA, the “upstream” or “5’ ” (“five-prime”) portion is shown on the left and the “downstream” or “3’ ” (“three-prime”) portion is shown on the right, with the encoding sequence in the middle. (Id.) One can think of this as a section of rope with a knot on one end signifying the 5' end of the locus, the middle section as a gene segment, and a knot on the far end as the 3' end of the locus. Thus, the immu-noglobulin loci have a 5' end and a 3' end; in between are the heavy and light chain variable region gene segments and the heavy and light chain constant region segments, with the variable segments arranged at the 5' end and the constant segments toward the 3' end. One can think of the 5' and 3' ends as the boundary lines of the locus. Outside of the 5' and 3' one is outside of the locus; inside the boundary lines are all of the various regions including the variable regions (heavy and light) and constant regions (heavy and light). Thus, knowing the 5' and 3' defines the playing field — but where on the playing field one desires to place the “ball” (or gene segment), if one desires a specific location, requires additional information. Continuing with our playing field analogy, to place the ball at the 50 yard line, one needs to know where that is. Of course, one also needs to know whether the coach cares where on the field the ball is placed, or whether the intent is just to get it onto the field. The “coach” (scientist) may be indifferent. This concept is important for the invention at issue in the '018 Patent — both because Merus claims that the metes and bounds of the 5' was unknown in February 2001, and because Regeneron now claims that the location within the locus (in our analogy, the precise point on the playing field) at which the targeting needs to occur, is quite specific. As discussed below, Regeneron asserts (through Dr. Oettinger) that the insertion of the human DNA segment must occur distal to, and upstream of, the 5' such that it is next to, but not within, the area which contains the mouse constant regions. Thus, performing targeted insertion of variable gene segments into the immuno-globulin locus of a mouse requires choosing a homology arm upstream (5’) of the chromosomal fragment, and a homology arm downstream (3’) of that fragment. If homologous recombination occurs within the boundaries established by the homology arms, the insertion has been accomplished. Figure 4 is worth repeating here to illustrate this: Notice that in the figure above, depicting an example of insertion, the transgene (that is, the gene from the outside human organism) has been added into, but has not replaced, the existing genes in the mouse, As discussed, a separate process would be required to delete the pre-existing homologous segment, or to inactivate it. Over time, scientists skilled in the art have found that human gene segments inserted into a mouse genome, and into the immunoglobulin gene in particular, are able to rearrange and thereby produce a broad spectrum of VDJ and VJ regions (for heavy and light chains) that are expressed in antibodies. The method set forth in the '018 Patent may result in genetically modified mice that can produce antibodies useful in drug discovery and downstream production of potentially useful therapeutic antibodies. (Davis Tr. Decl. ¶ 17.) As discussed below, that same method may also, however, produce a mouse capable of producing inferior or even useless antibodies. This might occur if (1) the insertion occurs in the gene (e.g. somewhere on the playing field) at a point that is not next to the constant region — perhaps even within the constant region; (2) the inserted human gene segment is only one portion of the variable region (e.g. heavy but not light chain or vice versa), or (3) the homologous mouse gene segment is not deleted or inactivated, but instead continues to exist within the locus. It should be clear by this point that Drs. Oettinger and Davis do not agree on various technical aspects of the invention. Indeed, they disagree on certain fundamental points. In making its determinations herein, the Court has read the material submitted by each and had the opportunity to see them on cross-examination and redirect, and also to pose certain questions itself. While the Court does find Dr. Oet-tinger experienced in the relevant area, the Court credits Dr. Davis’s views on technical aspects in which they differ, including on the invention. That is due to the reasoned basis for Dr. Davis’s views, the evidence he brought to bear to support his views, and how he responded on cross examination. As stated above, the Court’s technical statements are, therefore, findings of fact. Among their disagreements is whether insertion of the human V-D-J/V-J (that is, both heavy and light chain variable regions) and deletion of the homologous mouse sequences leaves intact all of the sequences, including regulatory sequences, necessary to enable the production of useful antibodies. Dr. Oettinger asserts that it does, and Dr. Davis disagrees. Dr. Davis testified credibly that “[t]here are ... important differences between the loci in organization, regulation and existence of embedded genes associated with other functions in the organism, which do not comport with the assertion” that all necessary sequences for proper transcription, recombination and/or class switching are left intact. (Id. ¶ 27.) One example highlights the importance of this disagreement. Following its submission of the '018 Patent Application, Regeneron learned that embedded in the mouse heavy chain Ig locus there are genes (referred to as ADAM6) that are important for mouse fertility. If those genes are deleted according to the instructions set forth in the '018 Patent (and as referenced in Figures 4A-D of the '018 Patent), the resulting mouse will be infertile or have impaired fertility. (Id. ¶ 28; DX 159, at 734, 737; see also DX 3, U.S. Patent No. 8,642,835 at 1:15-28.) Another point of difference concerns the identity of light and heavy chains. For instance, Dr. Davis takes issue with Dr. Oettinger’s assertion that a naturally occurring Ig molecule always has two identical light chains and two identical heavy chains. Dr. Davis states, with support, that (for example) the IgG4 is “inherently unstable” and “exchange of HL pairs may occur resulting in different heavy chains and/or light chains in the circulation.” (Davis Tr. Decl. ¶ 25.) He further states, with support, that while Dr. Oettinger asserts that the unrearranged variable region gene structures of heavy and light chains are similar, they are not always so. (Id. ¶ 27.) For instance, mouse heavy and light chain Ig loci organization and content are different. “[T]he mouse endogenous lambda locus has regulatory elements, and constant regions sandwiched between V and J gene segments.” (Id. ¶ 29.) According to Dr. Davis, and the Court credits his testimony in this regard, if one skilled in the art were to follow the targeting strategy set forth in the '018 Patent for the lambda locus, he or she would be removing mouse constant regions and regulatory elements within that locus. (Id. ¶ 30.) This conflicts with Dr. Oettinger’s assertion that the invention requires maintaining the totality of the mouse constant region intact. (Id. ¶ 31.) III. PROSECUTION HISTORY OF THE '018 PATENT U.S. Patent Application No. 13/164,176 (the '176 Application), entitled “Method of Modifying Eukaryotic Cells,” was filed on June 20, 2011. (See '018 Patent.) The application issued as U.S. Patent No. 8,502,018 (the '018 Patent) on August 6, 2013, to inventors Drs. Andrew J. Murphy and George D. Yancopoulos, (Id.) and was assigned to Regeneron. As originally filed, claim 1 of the '176 Application describes “A genetically modified mouse, comprising in its germline human unrearranged variable gene region segments inserted at a mouse immunoglo-bulin locus.” (DX 2 at 44.) But for the later inclusion of the word “endogenous”, this is identical to claim I of the '018 Patent as issued. On January 26, 2012, the PTO issued a Non-Final Office Action rejecting claims 1-19 of the '176 Application as being anticipated by a Lonberg reference, 2006/0015957 (Id. at 128-39.) That Office Action stated, in part: Lonberg and Kay teach heterologous unrearranged immunoglobulin human heavy and light chain transgenes useful for producing transgenic mice ... and transgenes are typically integrated into host chromosomal DNA, into germline DNA. Lonberg and Kay teach the production of chimeric human variable region/mouse constant region antibodies through trans-switching ... thus the mouse does not comprise a human im-munoglobulin constant region gene. (Id. at 131-32.) On July 26,. 2012, Regeneron’s Dr. Tor Smeland, in-house counsel responsible for prosecuting that application and others in the same family in the United States and Europe, replied to this Office Action. He argued, inter alia, that unlike the '176 Application, .Lonberg teaches random and not targeted insertion: Lonberg does not disclose a mouse comprising in its germline human unre-arranged variable region gene segments inserted at a mouse immunoglo-bulin locus. Instead, Lonberg discloses transgenes that are apparently randomly inserted at (unknown) loci. Lonberg simply lacks description of the claimed chimeric locus of claim 1. Amended claim 11 and amended claim 20 also recite a chimeric endogenous locus, which is not disclosed in Lonberg. Thus, regardless of whether Lonberg disclosed chimeric human variable/mouse constant antibody proteins, Lonberg does not anticipate the claims because a disclosure of trans-switching does not disclose ... endogenous mouse loci that are modified as claimed ... The claimed method does not represent a selection from predictable solutions, i.e., the claimed method was not “obvious to try” at the time it was filed. An obvious to try argument assumes a design need or market pressure to solve a recognized problem in order to achieve an anticipated success. The art never recognized (1) that there was a “problem” to be solved in making antibodies from an endogenous mouse locus, or (2) that there was a design need or market pressure to achieve success at modifying an endogenous mouse immunoglobulin locus to make a chimeric endogenous locus. (Id. at 160-61, 163 (emphasis added).) On October 11, 2012, the PTO mailed a Final Office Action, rejecting the pending claims of the '176 Application. The Final Office Action maintained the rejection, of claims 1-19 as anticipated by Lonberg. (Id. at 180.) In a January 11, 2013 Reply to the Final Office Action, Regeneron amended claim 1 to include the additional limitation that the human unrearranged variable region gene segments would be inserted at “an endogenous” mouse immunoglobulin locus. (Id. at 202.) In connection with that amendment Regeneron stated: The Lonberg paragraphs cited by the Examiner merely disclose that human transgenes for making human antibodies were mentioned in the art. None of the cited paragraphs suggest or even hint at placing unrearranged human immuno-globulin gene segments at an endogenous mouse locus, much less a functional endogenous mouse locus. The cited portions of Lonberg leave no doubt whatsoever that the Lonberg mouse construction instructions were to build a transgenic mouse that makes fully human antibodies from transgenes that are distant from endogenous mouse immu-noglobulin loci; i.e., they are synthetic loci randomly inserted into the mouse genome at a locus distant from any functional mouse immunoglobulin locus. Indeed, as is described in detail elsewhere in Lonberg, the Lonberg transgenic mouse requires that endogenous mouse immunoglobulin loci (both heavy and light chain loci) must be rendered nonfunctional so as to allow the fully human immunoglobulin transgenes to make fully human antibodies. There is absolutely no hint or suggestion in Lonberg to employ a functional endogenous mouse locus having inserted unrearranged human immunoglobulin variable region gene segments in the functional locus. (Id. at 204-05.), The reply also represented that the Vel-oclmmune mouse is the commercial embodiment of the invention: However, regardless of whether the Examiner has made a prima facie case of obviousness with respect to claim 20, Applicants submit that claim 20 is patentable because the claimed mouse exhibits features entirely unexpected in lights of the teachings of prior art (e.g., Lonberg, Brüggemann, Kawasaki, and Popov). The features of mice having disabled endogenous immunoglobulin loci and comprise transgenes that make antibodies with human variable domains have been .disclosed in peer-reviewed publications disclosed in the information disclosure statement filed in this application, dated 20 September 2011. The claimed mice, an embodiment of which is known in the art as a VELO-CIMMUNE humanized mouse, perform surprisingly and unexpectedly better than mice with disabled endogenous loci that express antibodies from randomly inserted transgenes (as in all of the references cited by the Examiner). (Id. at 209 (emphasis added).) Attached to Regeneron’s reply was a slide presentation (id. at 214-32) that Dr. Smeland provided to the PTO, and which he and Brendan Jones, an outside patent attorney retained to represent Regeneron in the final stages of prosecution of the Patent, relied on in a meeting with the PTO. (See id. at 290.) That presentation contains information which Merus asserts is false and was known to be false at the time. It concerns the Veloclmmune mouse to which Dr. Smeland’s January 2013 reply referred. Various figures in that presentation describe ways in which the Veloclmmune mouse was made. These figures are consistent with the presentation’s assertion that the Veloclmmune mouse was “Created only by virtue of Vel-ociGene & VelociMouse technologies.” (Id. at 215.) As discussed more fully below, this Court agrees with Merus that these slides provide certain misleading and inaccurate information. First, as of February 2001, the Veloclmmune mouse did not exist— Regeneron had been unable to make it. (See, e.g., DX 145; REGN-AM-10055694.) Yet the presentation suggests it did. In addition, on slide 10, a figure depicts the locus construction for the Velo-clmmune mouse. It indicates that Regen-eron replaced a 3 mb segment with a 150 kb segment in a single step; that is, that both insertion and deletion occurred simultaneously. (DX 2 at 224.) This was not in fact the pi’ocess used to produce the Velo-clmmune mouse. (Davis Tr. Decl. ¶ 279.) As discussed above, to insert both human heavy and light chain variable regions requires two steps (or a breeding step), and a third step is required to delete or inactivate the homologous mouse sequence in order to obtain therapeutically useful antibodies. Moreover, in February 2001 (and for a substantial number of years thereafter), Regeneron had not succeeded in inserting and deleting a portion of mouse IgH DNA that was over 200 kb. (See, e.g., DX 145; REGN-AM-10055694.) Nevertheless, the '018 Patent depicts this in Figure 4 and the presentation indicates that insertion and deletion on this scale had occurred. Figure 4 of the '018 Patent shows a replacement of approximately 200-300 kb of human immunoglobulin DNA for mouse immunoglobulin DNA. (’018 Patent at Fig. 4). In addition, the presentation discusses the ability of the Veloclmmune mouse to preserve the transmembrane and cytoplasmic DNA of the endogenous mouse immunoglobulin locus as among its benefits over prior art mice. (DX 2 at 219, 222.) The presentation discusses the preservation of these regions as the “Velo-clmmune Hypothesis.” (Id. at 226.) But neither the claims nor the specification contains such a limitation. (See '018 Patent, 3:27-8:3, 29:24-30:64.) Moreover, this concept was not novel. One of the references Regeneron had not disclosed to the PTO (and at issue in this proceeding), Zou, in 1994 disclosed the preservation of mouse constant cytoplasmic and tran-smembrane domains and stated that the mice produced humanized antibodies “at the same level and efficiency as wild-type mice produce murine IgGl antibodies.” (DX 72, Zou, et al. (1994) at 1099.) These undisclosed results undercut the claims of the Veloelmmune mouse’s superiority found in Dr. Smeland’s January 2013 presentation, which extolled “Normal variable region usage and junctional diversity,” as well as “Normal numbers and distribution of B cells in spleen and lymph node” and “Normal B cell differentiation in bone marrow.” (DX 2 at 227; Davis Tr. Decl. ¶ 349.) Dr. Andrew Murphy of Regeneron was one of the authors (but not presenters) of the slides that were provided to the PTO during patent prosecution. Prior to creating the January 2013 slide deck, Dr. Murphy had been told by another pioneering scientist in the field who had been on Regeneron’s Scientific Advisory Board, Dr. Frederick W. Alt, that assertions that Vel-oclmmune mice demonstrated no major defects in B cell differentiation “could be a little misleading.” (DX 223 at 10039849; DX 111 REGN-AM-00061940) Dr. Alt shared this comment in an August 15, 2011 message that provided comments on a manuscript Dr. Murphy had sent Dr. Alt and others the prior March. In the March email, which was titled “Veloelmmune manuscripts,” Dr. Murphy had told the recipients they were “listed as a co-author in one or both of the enclosed manuscripts,” and asked for any edits. (DX 112.) In his comments on August 15, 2011, Dr. Alt responded to an assertion in the manuscript that read “No major defects were observed in B cell differentiation in any of the Veloelmmune mice. The introduction of human IgH variable segments does not appear to affect either the pro B to pre-B transition nor do human IgK variables affect the proB to B transition.” (DX 223 at 10039848.) Dr. Alt wrote that, in his view, this statement was “correct but perhaps could be a little misleading.” (Id. at 10039849.) He explained that when we looked at bone marrow BM there was a profound block in the pro-B and pre-B transition, suggesting that there is significant selection/expansion of the 3 human VH locus to get a normal percentage of B cells in the periphery. ... [I]n reality if you have too few human VH then you may have impaired development and therefore the number of VHs is important, but once you have a certain number of VH genes (for example 18 in Velcoimmune), there is no obvious developmental impairment.” (Id.) Another recipient of that same email, Dr. Klaus Rajewsky, also provided comments to Dr. Murphy. He advised Dr. Murphy that “[sjince the first paper deals in depth with the issue of replacing mouse by human immunoglobulin gene segments, it may be appropriate to quote the first paper(s) demonstrating such replacements, which were actually done in my lab almost 20 years ago. The references are attached.” (DX 113.) One of the attached references was the Zou reference that is alleged to be one of the Withheld References in this proceeding. Having received this information from both Drs. Alt and Rajewsky, and without any evidence in the record suggesting his colleagues’ comments were unfounded or incorrect, Dr. Murphy nevertheless assisted in authoring the presentation to the PTO that continued to assert that the Vel-oclmmune mouse with 3 VH gene segments was “normal” meaning “identical to wild-type mouse littermates,” ignoring Dr. Rajewsky’s prior lab work and the Zou publications. (DX 2 at 227.) Following receipt of the January 2013 presentation from Dr. Smeland, the PTO issued an Advisory Action maintaining the rejection of claims 1-19 as anticipated by Lonberg, and claim 20 remained rejected in view of Lonberg and other references. {Id. at 241, 248.) Shortly thereafter, on February 19, 2013, Regeneron retained Brendan Jones, Ph.D., to assist with prosecution. (Id. at 268.) Drs. Jones and Sme-land together planned an in-person meeting with the PTO at which Regeneron relied on the previously provided slide deck described above. That meeting occurred on March 11, 2013. (Id. at 290.) Following that meeting, the Examiner prepared the following notes: “Applicant’s representatives discussed that Lonberg does not teach integration of human unre-arranged immunoglobulin genes into an endogenous site of a mouse immunoglobu-lin locus as required by the instant claims.” (Id.) The Examiner agreed to review the pending application. (Id. at 301.) On April 26, 2013, the PTO issued a Notice of Allowance for the '176 Application. (Id. at 285.) In the statement of reasons for allowance, the Examiner stated that “[t]he prior art does not teach or suggest a genetically modified mouse comprising, in its germline cells, human unre-arranged variable region gene segments inserted at an endogenous mouse immuno-globulin locus.” (Id. at 283; ECF No. 241 ¶ 172.) The applicant transmitted the fee on June 28, 2013 and the patent issued as the '018 Patent on August 6, 2013. (DX 2 at 328-29, 339; '018 Patent.) IV. LEGAL STANDARDS FOR A FINDING OF INEQUITABLE CONDUCT Merus asserts that Drs. Smeland and Murphy violated their duty of candor and engaged in inequitable conduct. Merus also alleges that Drs. Smeland and Murphy engaged in egregious affirmative misconduct by, inter alia, making false and misleading statements and including false and misleading results in the January 2013 presentation. Regeneron does not contest that both of these individuals had a duty of candor to the PTO, but vigorously contests whether that duty was violated, whether any non-disclosure rose to the level of inequitable conduct as defined by Thera-sense, and whether either Drs. Smeland or Murphy engaged in egregious misconduct. Each individual associated with the prosecution of a patent has a duty of candor and good faith to the PTO. 37 C.F.R. § 1.56(a). This duty includes a “duty to disclose to the Office all information known to that individual to be material to patenta-bility ...” Id. The doctrine of inequitable conduct — which can render a patent unenforceable — has origins in those duties as well as a lengthy body of caselaw. In 2011, the Federal Circuit made it clear, however, that the statutory duties of candor and disclosure and the caselaw doctrine of “inequitable conduct” are not always coextensive. See Therasense, Inc. v. Becton, Dickinson & Co., 649 F.3d 1276, 1291-1292 (Fed.Cir.2011) (en banc). “As an equitable doctrine, inequitable conduct hinges on basic fairness.” Id. at 1292. “[A]s a general rule, this doctrine should only be applied in instances where the patentee’s misconduct resulted in the unfair benefit of receiving an unwarranted claim.” Id. (citing Star Sci., Inc. v. R.J. Reynolds Tobacco Co., 537 F.3d 1357, 1366 (Fed.Cir.2008)). The Federal Circuit’s en banc decision in Therasense sets forth the governing legal standard. After noting that asserting claims of inequitable conduct had “become a significant litigation strategy” that can “cast a dark cloud over a patent’s validity and paint the patentee as a bad actor” and increase the costs and complexity of infringement litigation, id. at 1288, the Court proceeded to “tighten]] the standards for finding both intent and materiality in order to redirect a doctrine that has been overused to the detriment of the public.” Id. at 1290. A court’s determination of inequitable conduct proceeds in two parts: the accused infringer, who bears the burden of proof on this claim, must prove both that a nondisclosed reference was material and that the patent applicant acted with the requisite intent. See id. “[A]s a general matter, the materiality required to establish inequitable conduct is but-for materiality. When an applicant fails to disclose prior art to the PTO, that prior art is but-for material if the PTO would not have allowed a claim had it been aware of the undisclosed prior art.” Id. at 1291. The Court is. therefore required to place itself in the shoes of a patent examiner and determine whether it would have allowed the claim “if it had been aware of the undisclosed reference.” Id. In making its determination as to materiality, “the court should apply the preponderance of the evidence standard and give claims their broadest reasonable construction.” Id. at 1291-92 (citing Manual of Patent Examining Procedure (“MPEP”) §§ 706, 2111 (8th ed. Rev. 8, July 2010)). Whether prior art is material is determined by one with ordinary skill in the art. Air-Site Corp. v. VSI Int’l, Inc., 174 F.3d 1308, 1324 (Fed.Cir.1999). A court can take into account the inferences and creative steps that a person of ordinary skill in the art would employ when deciding whether a claimed combination of prior art would render an invention obvious. DyStar Textilfarben GmbH v. C.H. Patrick Co., 464 F.3d 1356, 1366-68 (Fed.Cir.2006). A finding by a district court that withheld information, such as a prior art reference, renders one or more claims invalid (for instance, by rendering it obvious or anticipated), indicates that the reference is necessarily but-for material. Aventis Pharma S.A. v. Hospira, Inc., 675 F.3d 1324, 1334 (Fed.Cir.2012); Am. Calcar, Inc. v. Am. Honda Motor Co., 651 F.3d 1318, 1334 (Fed.Cir.2011); Therasense, 649 F.3d at 1292 (finding reference was “necessarily material” where the jury and court found reference anticipated asserted claims). Of particular importance here is the treatment of prior art in connection with other related patent applications. Rejections over withheld prior art in other patent applications with similar claims is evidence of. materiality. See Dayco Prods., Inc. v. Total Containment, Inc., 329 F.3d 1358, 1368 (Fed.Cir.2003) (“We hold that a contrary decision by another examiner reviewing a substantially similar claim meets the Akron Polymer [Container Corp. v. Exxel Container, Inc., 148 F.3d 1380 (Fed.Cir.1998) ] ‘reasonable examiner’ threshold materiality test of ‘any information that a reasonable examiner would substantially likely consider important in deciding whether to allow an application to issue as a patent.’... A prior rejection of a substantially similar claim refutes, or is inconsistent with the position that those claims are patentable. An adverse decision by another examiner, therefore, meets the materiality standard under the amended Rule 56.”); see also, Larson Mfg. Co. of S. Dakota v. Aluminart Prods. Ltd., 559 F.3d 1317, 1339 (Fed.Cir.2009) (“Because the Third and Fourth Office Actions contained another examiner’s adverse decisions about substantially similar claims, and because the Third and Fourth Office Actions are not cumulative to the First and Second Office Actions, the district court correctly found the withheld Office Actions material.”); McKesson Info. Solutions, Inc. v. Bridge Med., Inc., 487 F.3d 897, 918-925 (Fed.Cir.2007). A reference is not but-for material if it is merely cumulative. See, e.g., Larson, 559 F.3d at 1331; McKesson, 487 F.3d at 913; Dig. Control Inc. v. Charles Mach. Works, 437 F.3d 1309, 1319 (Fed.Cir.2006) (“However, a withheld otherwise material prior art reference is not material for purposes of inequitable conduct if it is merely cumulative to that information considered by the examiner.”); Molins PLC v. Textron, Inc., 48 F.3d 1172, 1179 (Fed.Cir.1995); Litton Indus. Prods. Inc. v. Solid State Sys. Corp., 755 F.2d 158, 167 (Fed.Cir.1985). A reference is cumulative when it “teaches no more than what a reasonable examiner would consider to be taught by the prior art already before the PTO.” Regents of the Univ. of Calif. v. Eli Lilly & Co., 119 F.3d 1559, 1575 (Fed.Cir.1997). When a particular reference discloses a limitation of particular importance not elsewhere disclosed, it is not cumulative. McKesson, 487 F.3d at 914. Similarly, when a reference contains a more complete combination of the elements claimed, it is not cumulative even if the elements are before the examiner in other references. Semiconductor Energy Lab. Co. v. Samsung Elecs. Co., 204 F.3d 1368, 1374 (Fed.Cir.2000). Finally, the mere existence of differences between a withheld reference and the claims does not, alone, render the reference immaterial. See McKesson, 487 F.3d at 915 (citing Li Second Family Ltd. v. Toshiba Corp., 231 F.3d 1373, 1380 (Fed.Cir.2000)). Materiality and intent are separate, independent prongs of the inequitable conduct inquiry. Therasense, Inc. v. Becton, Dickinson & Co., 649 F.3d 1276, 1290 (Fed.Cir.2011) (en banc). The requisite specific intent to deceive must be proven by clear and convincing evidence. Id. “[A] court must weigh the evidence of intent to deceive independent of its analysis of materiality. Proving that the applicant knew of a reference, should have known of its materiality, and decided not to submit it to the PTO does not prove specific intent to deceive.” Id. (citing Star Sci., Inc. v. R.J. Reynolds Tobacco Co., 537 F.3d 1357, 1366 (Fed.Cir.2008)). “To prevail on a claim of inequitable conduct, the accused infringer must prove that the pat-entee acted with the specific intent to deceive the PTO.” Id. “In a case involving nondisclosure of information, clear and convincing evidence must show that the applicant made a deliberate decision to withhold a known material reference.” Id. (emphasis in original) (quoting Molins PLC v. Textron, Inc., 48 F.3d 1172, 1181 (Fed.Cir.1995)). The Court stated further, “[i]n other words, the accused infringer must prove by clear and convincing evidence that the applicant knew of the reference, knew that it was material, and mad